TFC Membrane

VANADIUM-REDOX FLOW AND LITHIUM-ION BATTERY
MODELLING AND PERFORMANCE IN WIND ENERGY

APPLICATIONS
By
John A. Chahwan
B.Eng. (McGill University, Montreal, Quebec)
Thesis submitted to the McGill University Department of Electrical and Computer

Engineering in partial fulfillment of the requirements of the degree of Master in
Engineering
Department of Electrical and Computer Engineering,

Mc Gill University,
Montréal, Québec, Canada
May 2007
© John A. Chahwan, 2007

1+1 Libraryand
Archives Canada
Bibliothèque et
Archives Canada
Published Heritage Direction du

Branch Patrimoine de l'édition
395 Wellington Street 395, rue Wellington
Ottawa ON K1A ON4 Ottawa ON K1A ON4
Canada Canada
Your file Votre référence

ISBN: 978-0-494-38480-0
Our file Notre référence
ISBN: 978-0-494-38480-0
NOTICE: AVIS:
The author has granted a non- L'auteur a accordé une licence non exclusive
exclusive license allowing Library permettant à la Bibliothèque et Archives
and Archives Canada to reproduce, Canada de reproduire, publier, archiver,
publish, archive, preserve, conserve, sauvegarder, conserver, transmettre au public
communicate to the public by par télécommunication ou par l'Internet, prêter,
telecommunication or on the Internet, distribuer et vendre des thèses partout dans
loan, distribute and sell theses le monde, à des fins commerciales ou autres,
worldwide, for commercial or non- sur support microforme, papier, électronique
commercial purposes, in microform, et/ou autres formats.
paper, electronic and/or any other
formats.
The author retains copyright L'auteur conserve la propriété du droit d'auteur

ownership and moral rights in et des droits moraux qui protège cette thèse.
this thesis. Neither the thesis Ni la thèse ni des extraits substantiels de
nor substantial extracts from it celle-ci ne doivent être imprimés ou autrement
may be printed or otherwise reproduits sans son autorisation.
reproduced without the author's
permission.
ln compliance with the Canadian Conformément à la loi canadienne

Privacy Act some supporting sur la protection de la vie privée,
forms may have been removed quelques formulaires secondaires
from this thesis. ont été enlevés de cette thèse.
While these forms may be included Bien que ces formulaires

in the document page count, aient inclus dans la pagination,
their removal does not represent il n'y aura aucun contenu manquant.
any loss of content from the
thesis.
•••
Canada
Abstract
As wind energy penetration levels increase, there is a growing interest in using

storage devices to aid in managing the fluctuations in wind turbine output power.
Vanadium-Redox batteries (VRB) and Lithium-Ion (Li-Ion) batteries are two emerging
technologies which can provide power smoothing in wind energy systems. However,
there is an apparent gap when it cornes to the data available regarding the design,
integration and operation of these batteries in wind systems. This thesis presents suitable
battery electrical models which will be used to assess system performance in wind energy
applications, inc1uding efficiency under various operating conditions, transfer
characteristics and transient operation. A design, sizing and testing methodology for
battery integration in converter based systems is presented. Recornrnendations for the
development of operating strategies are then provided based on the obtained results.
Résumé
La récente croissance en utilisation d'énergie éolienne augmente le besoin en

systèmes de stockage d'énergie pour diminuer les fluctuations de puissance produite par
le vent qui est de nature variable. Les batteries Vanadium-Redox (VRB) et Lithium-Ion
(Li-Ion) sont deux technologies émergentes qui peuvent être utilisées afin de réduire les
fluctuations éoliennes dans la production d'énergie. Cependant, l'information disponible
concernant la conception, l'implémentation et l'opération de ces types de systèmes est
insuffisante. Ce rapport présente une méthodologie pour la modélisation de batteries afin
de déterminer leur performance, incluant la fonction de transfert, le comportement
transitoire et le rendement. De plus, une méthodologie pour la conception et l'essai de
systèmes de batteries dans des systèmes d'éoliennes basés sur des convertisseurs statiques
est présentée. Des recommandations qui ont trait à l'optimisation des systèmes de
commande sont faites.
Acknowledgements
1 would like to sincerely thank my thesis supervisor and friend Dr. Géza JODS for his
guidance and support throughout my undergraduate and master' s studies at McGill
University. His insightful perspective on the engineering profession has helped me make
the right decisions with respect to my career and studies. Under his guidance, 1 was able
to challenge myself by setting higher objectives and fulfilling greater personal
achievements, while extending my technical knowledge.
1 would like to extend my thanks to Chad Abbey for his continued help and guidance
throughout my research project. 1 would also like to thank my peers in the Mc Gill
University power engineering department including Kyriakos Gogas, Jean Momeau, Wei
Li, Jose Restrepo, Xiaopeng Liu, Ping-Kwan Keung and Dr K. El Arroudi for forming a
strong and friendly power engineering community, as well as always being available to
help.
1 would also like to thank Melanie Chamberlain and Neeraj Gupta from NRC for their
appreciated contributions.
The financial support from "Le Fonds Québécois de la recherche sur la nature et les
technologies" is gratefully acknowledged.
Table of Contents
CHAPTER 1 INTRODUCTION •.•.•.•.•.•••••••••.•.•.••••••.•••••••••••••••..•.•••••••.•.••••••.••..•••••...•..•...•.•••• 1
1.1 Wind Energy Storage ........................................................................................... 1

1.2 Storage Research .................................................................................................. 2
1.2.1 Types ofStorage ......................................................................................................... 2
1.2.2 Common Batteries ...................................................................................................... 3
1.2.3 Wind Storage Research .............................................................................................. 4
1.3 Battery Modelling ................................................................................................ 5
1.3.1 Battery Performance................................................................................................... 5
1.3.2 Battery Modelling Overview ....................................................................................... 5
1.3.3 Model Evaluation ....................................................................................................... 6
1.3.4 Model Validation ........................................................................................................ 7
1.4 Existing Models ................................................................................................... 8
1.4.1 NiH2 Model ................................................................................................................ 8
1.4.2 Lead Acid Model ......................................................................................................... 8
1.4.3 Lithium-Ion Battery Models ........................................................................................ 9
1.4.4 VRB Models .............................................................................................................. 10
1.5 Research Objectives ........................................................................................... 12
1.5.1 Problem Definition ................................................................................................... 12
1.5.2 Identijj;ing the Gaps .................................................................................................. 12
1.5.3 Research Goals ......................................................................................................... 13
1.5.4 Claim of Originality.................................................................................................. 14
1.6 Thesis Outline .................................................................................................... 14
CHAPTER 2 MODEL DEVELOPMENT ...............••.............•..............•...•.........•........•....•.•• 16
2.1 Introduction ........................................................................................................ 16

2.2 Model Requirements .......................................................................................... 16
2.3 VRB Model Characteristics ............................................................................... 17
2.3.1 Model Specifications ................................................................................................. 17
2.3.2 InternaI Voltage and Temperature Considerations .................................................. 18
2.3.3 Rintemal and Rparasitic .................................................................................................... 19
2.3.4 State of Charge Modeling ......................................................................................... 20
2.3.5 Transient Operation ................................................................................................. 20
2.4 Lithium-Ion Battery Model Characteristics ....................................................... 21
2.4.1 Model Specifications ................................................................................................. 21
2.4.2 Internai Voltage and Losses ..................................................................................... 22
2.4.3 State of Charge Modeling ......................................................................................... 23
2.4.4 Transient Operation ................................................................................................. 23
2.5 SIMULINK Implementation .............................................................................. 24
2.6 Conclusions ........................................................................................................ 24
CHAPTER3 MODEL VALIDATION AND PERFORMANCE ............................................... 26
3.1 Introduction ........................................................................................................ 26

3.2 VRB Model Validation ...................................................................................... 26
3.2.1 Static Operation ........................................................................................................ 26
3.2.2 Dynamic Charge and Discharge Cycles ................................................................... 28
3.2.3 Energy Storage ......................................................................................................... 30
3.2.4 VRB Model Discussion ............................................................................................. 31
3.3 Li-Ion Model Validation .................................................................................... 32
3.3.1 Cel! Voltage .............................................................................................................. 32
3.3.2 Dynamic Charge and Discharge Cycles................................................................... 33
3.4 Standardized Profile for Experimental Tests ..................................................... 35
3.4.1 Typical Wind Profile ................................................................................................. 35
3.4.2 Charge-Discharge Profile ........................................................................................ 37
3.5 VRB Performance ......................... :.................................................................... 39
3.5.1 VI Transfer Characteristic........................................................................................ 39
3.5.2 System Efficiency ...................................................................................................... 39
3.5.3 Dynamic Response .................................................................................................... 43
3.6 Lithium-Ion Performance ................................................................................... 44
3.6.1 VI Transfer Characteristic........................................................................................ 44
3.6.2 System Efficiency ...................................................................................................... 44
3.6.3 Dynamic Response.................................................................................................... 47
3.7 Li-Ion and VRB Performance Comparison ...................................................... .47
3.8 Conclusions ........................................................................................................ 48
CHAPTER 4 WIND ENERGY STORAGE SYSTEM DESIGN ............................................... 49
4.1 Introduction ........................................................................................................ 49

4.2 Battery Sizing ..................................................................................................... 49
4.3 Scaling the VRB Model ..................................................................................... 51
4.3.1 System Requirements ................................................................................................ 51
4.3.2 Modeling a 42 kW Industrial VRB ............................................................................ 51
4.3.3 VRB Thevenin Equivalent ......................................................................................... 53
4.3.4 Multiple Stacks ......................................................................................................... 53
4.3.5 252 kW VRB System .................................................................................................. 55
4.4 Battery Converter Interface ................................................................................ 56
4.4.1 Buck-Boost de-de Converter ..................................................................................... 56
4.4.2 Inductance Design .................................................................................................... 56
4.4.3 Converter Controls ................................................................................................... 58
4.4.4 VRB de efficiency...................................................................................................... 59
4.5 Conclusions ........................................................................................................ 61
CHAPTER 5 WIND ENERGY STORAGE SYSTEM IMPLEMENTATION ............................ 63
5.1 Introduction ........................................................................................................ 63

5.2 Storage Schemes ................................................................................................ 63
5.3 DFIG Detailed System Model ........................................................................... 65
5.3.1 Model Operation without Storage ............................................................................ 65
5.3.2 Controlling the Battery System ................................................................................. 66
5.3.3 Wind System Operation ............................................................................................ 68
5.4 DFIG Average System Model ........................................................................... 71
5.4.1 Operation without Storage ............................................. ;......................................... 71
5.4.2 Controlling the Battery System ................................................................................. 71
5.4.3 System Operation ...................................................................................................... 73
5.5 DFIG Simple Model for Real-Time Simulation ................................................ 76
5.5.1 Centralized storage................................................................................................... 76
5.5.2 Real-Time System Implementation ........................................................................... 76
5.5.3 Experimental Setup ................................................................................................... 76
5.5.4 Conffiguration and Validation of Control Systems .................................................. 78
5.5.5 Experimental Results ................................................................................................ 78
5.6 DFIG System Performance with VRB storage .................................................. 83
5.6.1 Battery Energy Management .................................................................................... 83
5.6.2 System Overall Efficiency ......................................................................................... 83
5.6.3 Impact on Grid ......................................................................................................... 84
5.7 Conclusions ........................................................................................................ 84
CHAPTER 6 CONCLUSIONS AND FUTURE WORK •••••••••••••••••••••••••••••••••••••••••••••••••••••••••• 86
6.1 Summary ............................................................................................................ 86

6.1.1 Battery Models ......................................................................................... ................. 86
6.1.2 Integration into Wind Systems .................................................................................. 86
6.2 Conclusions ........................................................................................................ 87
6.2.1 Modelling.................................................................................................................. 87
6.2.2 Operating Considerations ........................................................................................ 88
6.2.3 Design Considerations ............................................................................................. 88
6.2.4 A Methodology for the Study ofWind Storage Systems ............................................ 90
6.3 Future Research ................................................................................................. 90
List of Figures and Tables
Table 1: Available Power Storage Technologies 2
Table 2: Main Battery Storage Technologies 3
Fig. 1.1: Li-Ion impedance model [25] 9
Fig. 1.2: Li-Ion dynamic model [27] 10
Fig. 1.3: VRB simple model [17] 12
Fig. 2.1: Proposed VRB model 18
Fig. 2.2: Proposed Li-Ion model 22
Fig. 2.3: VRB model in SIMULINK 25
Fig. 2.4: Li-Ion battery model in SIMULINK 25
Fig. 3.1: Static operation ofa 3.3kW VRB at fixed 20% SOC 27
Table 3: Comparison of Results for Static Simulation 27
Fig. 3.2: VRB full discharge cycle (Initially 80% SOC, Pout = 3.3 kW) 28
Fig. 3.3: VRB full charge cycle (Initially: 20% SOC, Pin= 3.3 kW) 29
Fig. 3.4: Complete cycle for 3.3 kW 3 hour VRB 30
Fig. 3.5: 3.3 kW VRB discharge at 69.9 A for 3 hours (209 Ah) 31
Fig. 3.6: Li-Ion battery cell voltage versus SOC 33
Fig. 3.7: 40 Ah Li-Ion battery discharge at 40 A for 1 hour (100% initial SOC) 34
Fig. 3.8: 40 Ah Li-Ion battery charge at 40 A for 1 hour (0% initial SOC) 34
Fig. 3.9: Typical wind profile for storage device 37
Fig. 3.10: Typical VRB output power profile 38
Fig. 3.11: 3.3 kW VRB VI characteristic 39
Fig. 3.12: 3.3 kW VRB charge efficiency versus input currents 40
Fig. 3.13: 3.3 kW VRB discharge efficiency versus input currents 41
Table 4: Charge Efficiency for Various Operating Conditions 42
Table 5: Discharge Efficiency for Various Operating Conditions 42
Fig. 3.14: Worse case charge to discharge transition (80 A currents) 43
Fig. 3.15: Li-Ion VI characteristic 44
Fig. 3.16: Li-Ion battery charge efficiency versus current 45
Fig. 3.17: Li-Ion battery discharge efficiency versus current 46
Fig. 3.l8: Worse case charge to discharge transition (Initially 50% SOC, 80 A currents) 46
Fig. 4.1: 42 kW VRB model VI characteristic (Various SOC) 52
Table 6: Battery Steady State Thevenin Equivalent 53
Fig. 4.2: Two VRB models in series 54
Fig. 4.3: 252 kW battery bank 56
Fig. 4.4: Dc chopper implementation for VRB 57
Fig. 4.5: Charging and discharging the 42 kW VRB using a dc chopper 58
Fig. 4.6: Dc-dc chopper PI control 59
Fig. 4.7: 42 kW VRB CUITent control using the dc chopper 59
Fig. 4.8: Discharge efficiency versus CUITent for VRB-Chopper 60
Fig. 4.9: Charge efficiency versus CUITent for VRB-Chopper 61
Fig. 5.1: Local DFIG storage system 64
Fig. 5.2: Centralized storage system 65
Fig. 5.3: DFIG system power reference controls 66
Fig. 5.4: Power output response (DFIG detailed model, 252 kW VRB) 67
Fig. 5.5: Power output response time (DFIG detailed model, 252kW VRB) 67
Fig. 5.6: Overall control system 68
Fig. 5.7: Power output after sudden wind drop (DFIG detailed model, 252 kW VRB) 69
Fig. 5.8: VRB system after sudden wind drop (DFIG detailed mode l, 252 kW VRB) 69
Fig. 5.9: Power output after sudden wind rise (DFIG detailed mode l, 252 kW VRB) 70
Fig. 5.10: VRB system after sudden wind rise (DFIG detailed model, 252 kW VRB) 70
Fig. 5.11: Power output response (DFIG average model, 252 kW VRB) 72
Fig. 5.12: Power output response time (DFIG average model, 252 kW VRB) 72
Fig. 5.13: Wind profile 73
Fig. 5.14: Turbine output power without storage (DFIG average mode l, 252 kW VRB) 73
Fig. 5.15: Turbine output power with storage (DPlG average model, 252 kW VRB) 74
Fig. 5.16: Battery SOC (DFIG average model, 252 kW VRB) 75
Fig. 5.17: Storage system power response (DFIG average mode l, 252 kW VRB) 75
Fig. 5.18: RTS experimental setup 77
Fig. 5.19: Storage system power reference (Centralized storage, RTS) 79
Fig. 5.20: Storage system response experimental results (Centralized storage, RTS) 79
Fig. 5.21: Turbine output power without storage (Centralized storage, RTS) 80
Fig. 5.22: Turbine output power with storage experimental results (Centralized storage, RTS) 80
Fig. 5.23: Storage system power flow experimental results (Centralized storage, RTS) 81
Fig. 5.24: Battery SOC (Centralized storage, RTS) 81
Fig. 5.25: Dc voltages experimental results (Centralized storage, RTS) 82
Fig. 5.26: Dc bus CUITent experimental results (Centralized storage, RTS) 82
Fig. 6.1: Methodology for the study and design of wind storage systems 91
Chapter 1: Introduction
Chapter 1
Introduction
1.1 Wind Energy Storage
The motivation for developing and implementing electrical equivalent models

for batteries cornes from an interest in studying their application in wind energy systems.
As wind energy penetration levels increase, there is a growing interest in using short and
long-term storage devices to aid in managing the fluctuations in wind turbine output
power. Advantages ofusing these devices include:
• Short-term (seconds), medium-term (minutes) and possibly long-term (hours)

management of wind power fluctuations;
• Smoothing of large power output swings from wind farms;
• Providing real and reactive power for local voltage support during transmission
system short-circuits and large transients;
• Meeting peak-Ioad demands without interruption and without Increase In

generation: maximizing wind plant capacity;
Wind is by nature unpredictable. However, In an electrical system, load

demands must be instantaneously met. Storage systems can act to smooth out wind
energy variations and play a role in stabilizing the wind farm and local grid. The
importance of detailed models can thus not be overemphasized, as they help to dictate
what is reasonable as well as the constraints imposed upon the operating strategy.
ln systems like Doubly-Fed Induction Generator (DFIG) wind turbines, energy

storage has been demonstrated to provide smooth output power by acting as a real power
source or sink. Furthermore, it provides improved transient and dynamic responses, and a
potential to aid frequency and voltage regulation [1], [2], [3], [4], [5], [6].
- 1-
Energy storage systems can be used within a remote off-grid power system as
weIl as grid connected wind parks. In remote hybrid systems, there is an interest in
increasing wind penetration, reducing the diesel fuel consumption costs, as weIl as
avoiding voltage and frequency variations [7]. In interconnected power system wind
parks, the focus is on improving power quality and stability for a larger power range [8].
1.2 Storage Research
1.2.1 Types of Storage
There are several ongoing studies related to the use of energy storage in power
systems. Many authors have seen new potential in short and long-term storage devices
based on the latest developments in storage technologies [9], [10], [11], [12], [13].
Table 1: Available Power Storage Technologies
STORAGE ROUND TIME

ADVANTAGES DISADVANTAGES
TECHNOLOGY TRIPEFF. SC ALE
High efficiency Large capital cost
Pumped Long term storage Significant land area Hours-
70-80%
Hydroelectric Widely used Specifie topography Days
Polluting (Uses fuel)

High reliability
Large capital cost Hours -
Compressed Air Long term storage 85%
Operating costs Days
High flexibility Efficiency dependent on

Non-polluting electrochemistry Minutes -
Batteries 60-80%
Quick response High capital cost Hours
Long cyclic life Short term storage

Seconds -
Super Capacitor Small volume Limited power range 86%
Minutes
Short term storage
Very high efficiency
Large capital cost Seconds -
Flywheel Commercial success 89%
Large size Minutes
Refrigeration needs
Quick response Large capital cost
SMES 21 % Seconds
Low overall efficiency
-2-
Storage options include Pumped Hydroelectric, Super-Capacitor Energy,

Compressed Air Energy, Battery Storage (Vanadium-Redox, Lead Acid, and Lithium-
Ion), Superconducting Magnetic Energy and Flywheel Energy. Table 1 summarizes the
characteristics of these technologies.
1.2.2 Common Batteries
There are several battery technologies that are available for energy storage [14],
[15], [16] as shown in Table 2 below. Batteries offer flexibility in terms of sizing, while
providing a quick response.
Table 2: Main Battery Storage Technologies
ROUND
STORAGE TIME
LIFETIME TRIP DC-DC COMMENTS
TECHNOLOGY SCALE
EFF.
Highly scalable, Cost effective,

Vanadium-Redox Minutes-
10.000 cycles 80% Medium and long term
Flow Battery Hours
applications
High cost, Limited to lower

Lithium-Ion Seconds-
3000 cycles Near 100% power range, Short term
Minutes
(Li-Ion) applications, Quick response
High cost and maintenance

Seconds-
Lead Acid 3000 cycles 85% requirement, Short term
Minutes
applications
Temperature sensitive,
Sodium-Sulfur Minutes-
2250 cycles 89% Combined use for power
(NaS) Hours
quality and peak shaving
Zinc-Bromine Minutes- Highly scalable, Medium and

> 5000 cycles 75%
Flow Battery Hours long term applications
-3-
1.2.3 Wind Storage Research
Vanadium-Redox Batteries (VRB) and Lithium-Ion (Li-Ion) batteries in

particular appear to have potential for wide-scale implementation in wind energy systems.
VRB technology has many advantages inc1uding operation over a wide range of
power outputs, high storage efficiency, rapid response, low maintenance costs and long
lifecyc1e [17], [18]. Operation under rapidly changing conditions is possible without
impact on efficiency, because the integrated pump ensures the availability of electrolyte
at all times near the electrodes. The State of Charge (SOC) of a VRB is also directly
dependent on the amount of charged electrolyte in the tanks, and is thus easy to track.
Furthermore, the power and voltage ranges of a VRB depend on the cell stack,
while the energy capacity depends on the tank size and volume of electrolyte [12]. This
independence between energy and power ratings provides high flexibility in terms of
design. These characteristics make VRB technology well suited for wind energy
applications. There are ongoing projects that look at the integration of VRB in wind
systems [12].
Large capacity Li-Ion batteries are also emerging in fields such as satellites,
space stations and launch vehic1es [19]. Recent research for the use of Li-Ion batteries as
backup sources in large scale power systems has also been undertaken [20]. Li-Ion
batteries have several characteristics that make them attractive for large power
applications, such as excellent high-rate discharge performance, operation through a wide
range of temperatures, easy charge controllability and low self-discharge. Additional
advantages inc1ude long life performance as weIl as small size.
In order to study various aspects of battery storage in wind energy applications,

fast and accurate battery models are needed. These models could then be used within
complete wind-energy system simulations in order to obtain valid results on their benefits
and limitations, from both technical and economical perspectives.
-4-
1.3 Battery Modelling
1.3.1 Battery Performance
Battery perfonnance needs to be evaluated when companng these vanous

technologies. The obtained data can then be used to detennine how each battery should be
operated, and whether it meets the design requirements. Battery perfonnance data [14],
[21], [22] includes the following:
• Dc charge and discharge efficiency: this corresponds to the efficiency from the dc
bus, and includes the dc-dc chopper which controls the battery.
• Ac charge and discharge efficiency: this corresponds to the efficiency from the ac
bus, through a dc-ac converter. Converter losses are usually 5% round-trip.
• Discharge time, energy capacity and energy density
• Transient response time
• VI transfer characteristic
• Rated and maximum power
• Cycle life and temperature effects
1.3.2 Battery Modelling Overview
Batteries have a series of characteristics that dictate their behaviour. These

physical characteristics would ideally all be taken into account in a perfect model. They
include: lifetime, capacity, stability, charge-discharge behaviour as well as efficiency of
the charge-discharge process. Several types of battery mode1s exist, for different
applications and with varying accuracy. These include physical, empirical, abstract and
mixed models [22].
-5-
Physical models are the most complex and accurate models. Due to the
complexity of such models, simulations can take hours to days, and are primarily used by
battery designers.
Empirical models such as Peukert' s law [22] approximate a certain behavior of

the battery based on a simple mathematical formula. Parameters are determined by fitting
the expression to available experimental data. They often have limited accuracy and
insight.
Mixed models use high-Ievel representation of battery. The analytical

expressions are based on physical laws and experimental data determines the parameters.
These models have comparable accuracy to the physical models and give a high insight
into the current state of a battery. What they lose in accuracy, they gain in computational
speed.
Abstract models provide an equivalent representation of the battery. The level of

analytical insight will depend on each model and the number of variables used, however
there is a tradeoff between accuracy and simulation speed. Examples of such models
include discrete-time models, stochastic models and electric circuit models.
1.3.3 Model Evaluation
A model is characterized by the following ([22], [23], [24], [25], and [26]):
• Type (physical, empirical, abstract, mixed)
• Variables of interest and total number of parameters
• Computational complexity and simulation time
• Accuracy, analytical insight and assumptions
• In-depth battery knowledge requirement (chemistry)
• Performance and apparent tradeoffs
-6-
Therefore, a mode!' s pertinence can be validated with respect to a proposed

application, based on these characteristics. In general, an ideal model should allow the
optimum representation of real battery characteristics with a minimum set of parameters
1.3.4 Model Validation
Battery models need to be validated. This is accompli shed by evaluating the

simulation results of a model against measured experimental data. The scope of the
experimental data required is thus determined by what the proposed model claims to
represent.
For battery model validation, we are interested in a series of current and voltage
measurements. The data required for mode! validation, based on models reviewed ([17],
[22], [23], [24], [25] and [26]) can include the following:
• Voltage versus capacity at different temperatures
• Cycle life (discharge capacity versus cycle number)
• Load change transient response
• Performance under demanding conditions
• Input charging currents for varying available power
• Pulse and frequency response measurements
• Tafe! voltage curves (voltage versus current for several SOC) [26]
• Battery voltage dynamics (fast switch from charge to discharge)
• ,Constant current discharge and discharge idle time
• Constant CUITent charge, constant voltage charge and charge idle time
-7-
1.4 Existing Models
In order to develop accurate battery models that meet the requirements for wind
energy applications, ongoing research in battery modelling will tirst be reviewed in this
section.
1.4.1 NiH2 Model
The Nickel-Hydrogen (NiH2) battery complex electrochemical dynamic model

represents dc non-linear behavior and transient dynamics [26]. The math model consists
of the SUffi of two electrochemical process dc currents, which depend on the battery
terminal voltage. These are the Nickel plate charging current and the parasitic overcharge
reaction current.
This model is useful for dc analysis, low and mid-frequency stability and
dynamic analysis (dynamic transient effects). The dynamic elements and electrochemical
model parameters are all handled as programmable variables. Good model accuracy is
demonstrated using pulse and frequency response measurements, Tafel voltage curves
[26] and battery voltage dynamics.
1.4.2 Lead Acid Model
The mathematical Lead-Acid battery model [18] takes into account self-
discharge, battery storage capacity, internaI resistance, over-voltage and environmental
temperature. The model is tested and validated experimentally. The system consists of a
data acquisition system, a controller rack, and two temperature-controlled chambers.
Current controlled sources are used for testing.
The tests are done at different temperatures, voltage ranges as well as rates of
charge and discharge. The tests include constant CUITent discharge, dis charge idle time,
constant current charge, constant voltage charge and charge idle time. The experimental
results are closely correlated with simulation results.
-8-
1.4.3 Lithium-Ion Battery Models
The Lithium-Ion battery is one type of storage device of high interest for wind
energy systems. It appears to be weIl suited for small wind installations, such as the ones
found in remote cornrnunities.
The PSPICE macro-model [24] incorporates cell voltage, state of charge, rate
dependence and thermal characteristics. The model is developed based on ICR-18650
Lithium-Ion cells data. The pararneter coupling is dependent on lookup tables created
from data measurements. Battery characteristics modeled include state of charge,
degradation of cells over time, rate dependence, electrochemical delays, cell impedance
as well as thermal properties.
Model validation is accomplished by comparing measurement and simulation

curves. These curves include "voltage versus capacity" at different temperatures, cycle
life and load step transient response. The model is found to be accurate with errors of less
than 5% error for cycle life, and less than 12% for thermal performance. Load response
accuracy on the other hand could be improved by including more than one time constant.
The Li-Ion impedance model [25] on the other hand employs Electrochemical
Impedance Spectroscopy (EIS) to obtain physical-based non-linear equivalent circuit
models of super capacitors describing Li-Ion batteries. Voltage response and dc currents
are measured as ac current flows in the device, and the impedance spectrum is obtained.
The model, shown in figure 1.1, takes into account impedance dependence on
temperature and state of charge. Parameters are defined with respect to battery nominal
current and the number ofbattery cells connected in series.
Fig. 1.1: Li-Ion impedance model [25]
-9-
Comparing simulation results to measured data in the time domain validates the
model. Validation data also includes voltage and CUITent responses at certain SOC and
room temperatures. The accuracy obtained is excellent because Li-Ion batteries are suited
for this kind of model.
The dynamic lithium battery model [27], which is shown in figure 1.2,
represents the battery by a simple electrical circuit. The transients are represented by a
capacitor, while the losses are separated into two internaI resistances. The internaI voltage
source, representing the equilibrium potential, is expected to vary based on the SOc. The
model parameters are found by curve fitting, based on data obtained from manufacturers.
lo
Va --....;:c..: : : > - - - - , . - - - - ,
Fig. 1.2: Li-Ion dynamic mode) [27]
1.4.4 VRB Models
VRB technology uses a controlled pump to induce flow, which improves battery
performance and efficiency. These storage devices can be used with large wind
installations and other distributed generation, such as photovoltaic systems [28]. In
contrast with a Li-Ion battery, the main differences are:
• Large scale storage using electrolyte tanks
• Use of controller and pump for circulation
- 10 -
• Use of an ion exchange membrane, which can result in a larger fraction of the
internaI resistance being constant
• Independence between power and energy rating
• Longer lifetime
It has been shown experimentally that various VRB losses can be mode1ed using
constant resistive values [29]. The simple flow battery model [17], as shown in figure 1.3,
takes into account internaI resistance and parasitic resistance. The total energy storage of
the system depends on the SOC and amount of active chemicals in the system. The total
power available is re1ated to total electrode area within the cell stacks. The model also
accounts for system energy content and transient responses.
The internaI resistance accounts for losses due to reaction kinetics, mass
transport resistance, membrane resistance, solution resistance, electrode resistance and
bipolar plate resistance. Parasitic resistance accounts for power consumption by re-
circulation pumps, the system controller, and power loss from cell-stack by-pass currents.
In order to validate the model, various operating cases were examine d, induding
normal and extreme conditions. Sorne additional mode1 assumptions are made for the
pumping energy estimation:
• Pump and impeller efficiencies assumed constant
• System pressure drop increases with flow
• DC pump and simple resistive voltage controller
• Simple pump control algorithm
- 11 -
1
Re ontroller 1
1RFb<ed Losses
) CEleclrodes
Pump
L _ _ _ _'---_----.-----'
Fig. 1.3: VRB simple model [17]
1.5 Research Objectives
1.5.1 Problem Definition
The large-scale problem that needs to be solved is determining the technical and
economical advantages and limitations ofVRB and Li-Ion battery storage devices in wind
energy systems. The study of the appropriate battery models and their integration within
DFIG wind systems will provide the insight needed for system design assessment and
optimization.
1.5.2 Identifying the Gaps
There is an apparent gap when it cornes to the design, integration and operation
of large scale VRB and Li-Ion batteries in wind systems. Crucial information regarding
the batteries is not readily available, including efficiency under various operating
conditions, design and sizing methodologies as weIl as short, medium and long-term
operating strategies. This is due to an evolving technology and a competitive market in
which information is kept confidential by the manufacturers. Therefore, accurate generic
(not manufacturer specifie) models are thus needed, since they will provide generic
information without requiring on-site testing.
- 12 -
The purpose of this research project is thus to develop suitable benchmark VRB
and Li-Ion electrical models and to implement them into simulation software such as
SIMULINK. These models will be based on existing mathematical models that will be
modified and improved for the purpose of the research. The models will then be validated
mathematically and experimentally.
Wind energy storage system design also needs to be examined. The VRB model
will be integrated into a DFIG wind system, as well as incorporated into a real-time
OPAL-RT simulation. It will provide insight on how to design and operate the storage
system efficiently. These results will be used to demonstrate the pertinence of battery
storage integration in wind energy systems, as weIl as to provide recommendations with
respect to design and operating strategies.
1.5.3 Research Goals
The research goals are as follows:

1. Develop, improve, implement and validate electrical VRB and Lithium-Ion
battery models based on physical and mathematical principles.
2. Examine battery sizing issues and model scalability for large scale applications.
3. Develop a DFIG storage system by integrating a VRB within the wind system,
using simple, average and detailed DFIG models.
4. Develop areal-time simulation system using OPAL-RT, which combines real
and simulated components in order to experimentally validate the system design
methodology.
5. Develop a typical wind profile which can be used as a standard to
experimentally test and compare different batteries in wind systems.
6. Use the obtained performance data to draw conclusions on the operation of
battery storage systems within a wind energy context. Provide recommendations
to facilitate the design of short-term and long-term control algorithms.
- 13 -
1.5.4 Claim of Originality
To the best of the author's knowledge, the thesis provides new ideas in the field
of wind energy storage by:
• Developing and implementing an electrical VRB model with unique features

suitable for wind energy study.
• Developing and implementing an electrical Li-Ion battery model with unique

features suitable for wind energy study.
• Obtaining detailed performance data regarding the transfer characteristic, transient

operation and efficiency of VRB and Li-Ion battery technology under various
operating conditions.
• Demonstrating the operation of a DFIG-VRB system using the developed VRB

model within various DFIG systems (offline and real-time).
• Providing a methodology for the design of battery systems for wind energy
integration.
• Providing recommendations for the development of control algorithms.
• Developing a battery sizing approach.
• Developing a model scalability methodology.
1.6 Thesis Outline
In Chapter l, existing literature on the topics of wind energy storage and battery
modeling is reviewed. The role of storage in wind energy applications is examined. The
requirements for suitable models are then discussed.
Chapters II focuses on the development of suitable VRB and Li-Ion models. The
various elements that make up the models and their relation to the physical characteristics
of batteries are discussed.
- 14 -
In Chapter III, the models are validated through typical charge and discharge
simulations. They are used to obtain battery performance, including transfer
characteristic, efficiency under various operating conditions as weIl as transient operation.
VRB and Li-Ion battery performance is then compared. A tentative wind profile is also
presented, which can be used as a standard to testing wind-storage systems.
In Chapter IV, issues related to VRB design for a DFIG system are discussed.
This includes battery sizing, model scalability as weIl as battery interface. A dc-dc
converter interface is developed and used to control the battery. The impact of the
converter on battery performance is examined.
The VRB system is then integrated within a wind system in Chapter V. This is
done for both a detailed DFIG model, as weIl as an average DFIG model in SIMULINK.
The system is also implemented in a real-time experimental setup using OPAL-RT IIO
technology as weIl as real converters. The results show how the battery acts to smooth out
a DFIG wind turbine output under varying wind conditions.
Chapter VI concludes the thesis by reviewing the methodology developed, and

assessing the effectiveness of battery storage in wind energy systems. Recommendations
for the development of wind storage operating strategies are provided. This sets the
ground for future research on the topic of battery design and modeling in wind energy
applications.
- 15 -
Chapter 2: Model Development
Chapter 2
Model development
2.1 Introduction
It was determined that accurate e!ectrical models are needed for wind energy
storage system studies. In this chapter, mode! requirements are discussed. VRB and Li-
Ion models are presented. Various battery characteristics are examined, inc1uding the
power rating, cell stack voltage, pump losses, internaI and parasitic resistance losses, state
of charge representation and transients.
2.2 Model Requirements
The model needed requires specific characteristics. In terms of analytical

insight, it should pro duce results related to the storage efficiency of the process, transient
response and output transfer function. The model should consider how much energy can
be extracted given an amount of energy stored, and should inc1ude the impact on the
charge-discharge current. Furthermore, different levels of simulation time scales are
needed, and thus the mode! should be adaptable to various DFIG system mode!s.
In terms of complexity, the mode! must be accurate yet manageable in larger

system simulations. There must be a reasonable tradeoff between complexity and
computational time. In terms of scale, the mode! should be adaptable to typical power and
voltage levels encountered in various wind energy applications.
The battery capacity and discharge behavior need to be modeled appropriately

taking into account dependence on transient effects, temperature effects and the state of
charge (SOC).
- 16 -
2.3 VRB Model Characteristics
In a VRB, the total energy storage of the system depends on the SOC and
amount of active chemicals in the system. The total power available is related to the
electrode area within the cell stacks. The power and Energy are thus not related, and this
allows for easy scaling of the battery specifications as required.
The model to be implemented in SIMULINK takes into account prevlOUS

research related to VRB modeling [29], as well as VRB modeling data provided by
Natural Research Council Canada (NRC) [17]. It will improve on an existing
mathematical model, shown in figure 1.3, by filling the appropriate gaps. These gaps
include pump losses, SOC dynamics and scalability issues. The proposed model has the
following unique properties:
• The state of charge, which represents the amount of active chemicals in the
system, is modeled as a variable that is dynamically updated.
• The stack voltage is modeled as a controlled voltage source. The power flow
through this source impacts the changes in the SOC.
• The variable pump losses are modeled as a controlled current source.
The model also includes internaI resistance and parasitic resistance. The internaI
resistance accounts for losses inside the cell stack due to reaction kinetics, mass transport
resistance, membrane resistance, solution resistance, electrode resistance and bipolar plate
resistance. Parasitic resistance accounts for power consumption by re-circulation pumps,
the system controller, and power loss from cell-stack by-pass currents [17]. Figure 2.1
below shows the proposed model.
2.3.1 Model Specifications
The model to be implemented will be based on a 3.3 kW-3 hour 48 V residential

VRB system. The step-by-step procedure for deriving the parameters should be easily
repeatable for different sized VRB systems.
- 17 -
1 sact+! = sact + MaC 1
+ Parasitic losses
VStack
C Electrodes
RFixed Lasses
Ipump
1
+ Lasses
1
1
1
1
L ___________________________ J
RResistive Lasses
Fig. 2.1: Proposed VRB model
The calculations are based on estimating losses to be 15% internaI, and 6%

parasitic in the worse case operating point (end of cycle discharge), for a minimum
voltage of 42 V, and a CUITent of78.6 A. Thus for the battery to be able to provide 3.3 kW
with 21 % losses, the cell stack internaI power should be:
3300
P.tack = 1- 0.21 = 4177 W (1)
2.3.2 InternaI Voltage and Temperature Considerations
The battery individual cell voltage is directly related to the state of charge of the
battery based on the Nernst equation [17]:
~ell = "V"quilibrium + 2 RT ln( SOC )

FI-SOC
(2)
In the Nernst equation, Vequilibrium is defined as the cell potential at a SOC of

50%, F is defined as the Faraday constant equal to 96485 C/mole and R is defined as the
- 18 -
Chapter 2: Madel Development
universal gas constant equal to 8.314510 J/(K.moIe). The temperature impact on battery
operation is reflected by 'T'.
The internaI stack voltage (or open-circuit battery voltage) V Staek is modelled as
a controIled voltage source which depends on both the number of ceIls and the SOC. For
'n' ceIls, VStaek would be equal to n.Veell as shown in equation (3). In this case, 39 ceIl
stacks are needed.
V0t
"ac
k = n.(v .l.b.
eqUl' rlum
+ 2_R_T
F
In(_S_O_C_))
1 - SOC (3)
The battery output voltage V battery will thus depend on both the open circuit
voltage, as weIl as the operational losses. In steady state, if Istaek is defined as the input
stack cUITent, then V battery can be expressed as foIlows:
v;,attery = VStock + lstack'( Rreaction + RreSiStiVe _lasses) (4)
2.3.3 Rinternal and Rparasitic
TypicaIly, internaI resistance values for batteries are variable due to over-
potential issues [29]. However, in the case of a VRB, a larger fraction of internaI
resistance is constant. This aIlows estimating Rintemal to be constant. Rintemal is set to 0.101
Q, which is derived by estimating 15% losses at a maximum current of 78.6 A.
The parasitic losses are separated into fixed and variable losses [17]. The fixed
losses are represented as a fixed resistance and the variable losses as a controlled current
source. The losses are as follows:
PparaSitiC = P fued + k(~~~) = 84 + 42.s(~~~) (5)
The parasitic and pump losses are derived as follows:
(6)
- 19 -
Chapter 2: Madel Development
42.S( Istack )
SOC = 1.011( Istack ) (7)
42 SOC
The pump losses are modeled as a controlled CUITent source that is dependent on
the stack CUITent and state of charge, in parallel with the fixed parasitic resistance.
2.3.4 State of Charge Modeling
The system state of charge can be defined as:
Current Energy in Battery

State 0 f Charge = ------=------'- (8)
Total Energy Capacity
SIMULINK discrete simulations compute all variables at each time step.

Therefore one way to keep track of the state of charge is to update the SOC variable from
one time step to the next, based on the power that goes through the cell staçk. The SOC is
computed each cycle based on the previous SOC, using a fixed step simulation. The
change in SOC is modeled as follows:
(9)
MaC = _M
___ ~tack X TimeStep _ I stack x ~'tack X TimeStep (10)
Ecapacity EcapaCity P'ating x Timerating
2.3.5 Transient Operation
An important issue in battery modeling is transient behavior. The ability of the

system to respond quickly to fast changes is especially important for power smoothing
applications. In a VRB, the transient effects are re1ated to electrode capacitance, as well
as concentration depletion of active vanadium species near the felt fiber electrode.
The model will focus on transient behavior re1ated to electrode capacitance, as

shown in figure 2.1. Rintemal is divided into Rreaction (0.061 Q) and Rresistive losses (0.04 Q).
- 20-
Celectrodes is estimated to be 0.15 F for a 39 cell stack, where each cell has a 6 F series
capacitance. Concentration depletion is less of an issue for VRB technology, depending
on the pump system used.
2.4 Lithium-Ion Battery Model Characteristics
Li-Ion battery systems consists of a series of cells, each capable of storing a

fixed amount of energy. The energy capacity of a Li-Ion battery is expressed in Ah,
which corresponds to its discharge capacity at the rated current, for l-hour. The total
power available is thus dependent on the rated output voltage and current.
The model to be implemented in SIMULINK is based on previous research

related to Li-Ion battery mode1ing ([27], [30], and [31]). The model is based on an
existing mathematical model, shown in figure 1.2. Figure 2.2 below shows the proposed
model which takes into account the following Li-Ion properties:
• The State of Charge, which represents the amount of active chemicals in the
system, is modeled as a variable that is dynamically updated.
• The operating losses are modeled using resistances, separated into resistive and
reaction losses. This accounts for the finite conductivities of electrodes and
separators, concentration gradients of ionic species near the electrodes and limited
reaction rates at the electrode [27].
• The stack voltage is mode1ed as a controlled voltage source, dependent on the

number of cells and the SOC. The power flow through this source controls the
changes in the SOC.
• Transient effects are modeled using a capacitance across Rreaction.
2.4.1 Model Specifications
For comparison purposes with the VRB mode1, the Li-Ion battery mode1 to be
implemented will have roughly the same power rating. It will use 30 cells rated at 40 Ah,
which will pro duce an output power of 3300 W for an output voltage of 82.5 V. The
- 21 -
internaI voltage will vary between 81 V and 125 V depending on the SOC. Li-Ion
batteries are constructed using a series of cells which makes them easily scalable.
2.4.2 Internai Voltage and Losses
The battery internaI stack voltage varies between 2.7 V and 4.2 V. It is re1ated to
the state of charge of the battery based on the following formula:
v =V k 1( 0.9 x SOC ) (11)

cell equilibrium + . n 100 - 0.9 x SOC
(12)
This euqation is obtained by estimating the Li-Ion cell voltage characteristic

based on manufacturer data [27]. Vequilibrium was found to be 3.797. The k factor was
approximated to be 0.1829 for regular room temperature.
SOCI+! = SOC( + f).SOC 1
r
0--------------------(
+
1
VStack
CReaction
Fig. 2.2: Proposed Li-Ion model
- 22-
During steady state, V battery can be expressed as:
(13)
Based on typical manufacturer data, los ses of up to 5% can arise under worse
case conditions [30]. Thus RintemaJ will be estimated to account for 5% losses at a 40 A
rated cUITent, for a fully depleted battery. The internaI voltage in that case would be 81 V.
At a constant CUITent of a 40 A, an internaI resistance of 101 mn will produce 5% losses.
RintemaJ is then divided into two-thirds Rresistive and one-third Rreaction, which
amounts to 67.33 mn and 33.66 mn respectively. These values agree closely with the
ones proposed in [27].
2.4.3 State of Charge Modeling
Similarly to the VRB model, the state of charge is defined as:
Current Energy in Battery

State 0 f Charge = -----..:::.:....---~ (14)
Total Energy Capacity
The same approach is used to keep track of the SOC as was described in section
2.3.4: The SOC is computed each cycle based on the previous SOC, using a fixed step
simulation. The change in SOC for a Li-Ion battery is modeled as follows:
(15)
MDC = Ml. Ivtack x TimeStep (16)

Ecapacity I raling x Timerating
2.4.4 Transient Operation
The transient effects in Li-Ion batteries are related to double-layer formation at

the electrode/solution interface, capacitance due to purely electrical polarization and
capacitance from diffusion limited space charges. They will be approximated as a first
- 23 -
order transient response [27], and modeled using a single 4F capacitor per ceIl, across
Rreaction. This amounts to a total Creaction of 0.133 F for 30 ceIls in series.
2.5 SIMULINK Implementation
The developed models were implemented using SMULINK (SimPowerSystems

toolbox), as shown in figures 2.3 and 2.4. The following approach was used in both cases:
• The internaI battery voltage is implemented using a controIled voltage source

'Vec'. The voltage is controlled by the 'Stack Voltage' block.
• The 'Stack Voltage' block computes the battery voltage according to equations (3)
and (11). It takes into account the 'SOC' variable, as weIl as the number of ceIls
(set by the 'ceIls' constant).
• The 'SOC' block dynamically computes the SOC at each time step according to
equations (9) and (14). The change in energy content between each time step ~E
is computed as the product of the power flowing through 'Vec' by the time step.
• The battery total energy capacity is defined initially in the' SOC' block.
• The VRB and Li-Ion battery resistive and capacitive parameters are implemented
using elements from the SimPowerSystems library.
• In the VRB case, the pump losses current is implemented as a controllable current
source. The CUITent is computed according to equation (6).
2.6 Conclusions
Simple VRB and Li-Ion models were developed based on battery physical and
mathematical properties, as weIl as existing battery modelling data. They were
implemented in SIMULINK. They accounted for various battery characteristics including
operating losses, transient operation and energy storage. The next step is to validate the
models, and use them to obtain insight on battery performance including efficiency,
transfer characteristic and transient response.
- 24-
Istack 1+--------,
r-----------------------~SOC
Vstack 1+--------+--1 ~
~ '----:::s"'"Oc:::---'
~ Vec
l 1
,.~'oo 1~
tG
CELLS Velectrochemistry
C_Electrodes
Io--------I~SOC
StackVoltage ~IDC
?'"'"
~ '.'-'-'
Fig. 2.3: VRB model in SIMULINK
,...------------1 soc Istack 14--------,
soc Rresistive Rreaction
+ ~- -J\f\I'v-- -J\f\I'v--
IStack
Fig. 2.4: Li-Ion battery model in SIMULINK
- 25 -
Chapter 3: Model Validation and Performance
Chapter 3
Model Validation and Performance
3.1 Introduction
VRB and Li-Ion models which meet the project requirements were developed in
the previous section, based on battery physical and mathematical properties. Battery
models need to be validated against simulation and experimental results. A series of
possible tests was indicated in section 1.3.4. However, there are currently no published
standards or procedures for testing batteries specifically for wind energy applications.
Such data would include typical charge and discharge profiles that can be used to
compare various battery technologies.
In this chapter, the mode!s will be validated by running typical charge and
discharge simulations. Experimental validation is then discussed, and a standard wind
profile developed for testing. Finally, the models are used to obtain insight on the battery
performance, including efficiency, transfer characteristic and transient response. This data
will be used to compare VRB and Li-Ion technologies and their suitability for wind
energy applications.
3.2 VRB Model Validation
3.2.1 Static Operation
First, the mode! is validated by comparing simulated results against the results
presented in [17] for a state of charge of20%, and an output voltage of 43.5 V.
The obtained results can be se en in figure 3.1 and table 3. The largest error
obtained is of3%. Note that in order to be able to provide a rated output power of3313 W
we need an internaI stack power of 4270 W to compensate for operating losses.
- 26-
V
Out
-1Out
-
1Stack
v
+ Electrachemistry
R
Fi}(ed Lasses
Pump
Fig. 3.1: Static operation of a 3.3kW VRB at fixed 20% SOC
Table 3: Comparison of Results for Static Simulation
Parameter Results according to [11] SIMULINK simulation Error
Vstack 51.8 V 51.82 V >1%
Istack 81.9 A 82.4 A >1%
VinternaI 8.6V 8.32 V 3%
PintemaI 681 W 685.9 W >1%
Iparasitic 6.1 A 6.238 A 2%
Pparasitic 265W 271.3 W 2%
Iout 75.8A 76.2 A >1%
Pout 3300W 3313 W >1%
- 27-
3.2.2 Dynamic Charge and Discharge Cycles
The dynamic behaviour is examined by looking at the charging and discharging

of a 3.3 kW 3 hour (9.9 kWh) rated battery. In practice, operation is limited between 20%
and 80% SOC due to over voltage and under voltage issues.
3.2.2.1 Full Discharge Cycle
The discharging of a charged battery (SOC = 80%) at a constant output power of

3.3 kW is first considered. In this example, a dc CUITent source sets the output CUITent
based on the output voltage, keeping the output power at 3.3 kW. It is assumed that 9.9
kWh of stored energy can be used effectively. The simulation is executed in SIMULINK
for 3 hours, with a time step of 1second. The results can be seen in figure 3.2.
80 -=-------,---~---~---~---~-------,
~ 60
g 40
CI)
20L---~---L----L---~----L--~~
0.5 1 1.5 2 2.5 3
lime [hl
~ ::tL...---L-:~:~.~:~:1
0.5 1 1.5
lime [hl
2 2.5 3
~ ~~~---------------
è.. 48
Ë
> 44
46
: : :
42L----L---~---L---~----L----
0.5
•
1.5
lime [hl
2
•
2.5
J 3
Fig. 3.2: VRB full discharge cycle (Initially 80% SOC, Pout = 3.3 kW)
- 28-
80
~ 60
o 40
en
20~----~------~------~------~------~------=
0.5 1 1.5 2 2.5 3
lime [hl
0.5 1.5 1 2 2.5 3

lime [hl
> 52,------,------~------,_------~------,_----__,
:
g,
~ 48
>
~
ID
~
50t- - - - - -_ _ _ _ _ _ _ _ _ ___
III
46
44
: : :
42~----~------~------~------~------~----~
0.5 1.5
lime [hl
2 2.5
•
J 3
Fig. 3.3: VRB full charge cycle (Initially: 20% SOC, Pin= 3.3 kW)
The internaI discharge power is kept constant close to 4100 W. This ensures that
the output power will be 3.3 kW after losses. The voltage does in fact vary linearly in the
20% to 80% SOC. The battery model meets the requirements in terrns of discharge
voltage and SOC profiles.
3.2.2.2 Full Charge Cycle
The charging of a discharged battery (SOC = 20%) at a constant input power of

3.3 kW is then considered. The simulation is executed in SIMULINK for 4 hours, with a
time step of 1second. The results can be seen in figure 3.3.
The charging power is kept close to 3.3 kW. The voltage varies linearly as weIl
in the 20% to 80% SOC region. Due to the losses, the actual power getting stored varies
between 2850 W and 2700 W at that CUITent. Therefore, the battery must be charged at
3.3 kW for more than 4 hours to reach 80% SOC. The battery model does however meet
the requirements in terrns of charge voltage and SOC profile.
- 29-
3.2.2.3 Full Cycle at Constant Current
Figure 3.4 shows a full charge-discharge cycle. The voltage and battery output
power can be observed as the SOC varies between 10% and 90%. The charging CUITent of
65 A is followed by a discharging CUITent of 65 A.
(ft. 100
Ü 50
oCI)
O~--~--~----~--~----~--~----~--~----~~
o 1 234 567 8 9
Time [hl
i~~r : : : : l : : : :
~ 0L---~--~2----~3----~4----~5----~6----~7----~8----~9~
Time [hl
J
~ 70~--~--~----,----,----,----,----,----.----,-,
!:t'- - - - -' - -:- '- - - :- - - :~:~ - '- - - -:'- :~;J

~ 0 2 3 4 5
Time [hl
6 7 8 9
Fig. 3.4: Complete cycle for 3.3 kW 3 hour VRB

10 to 90% SOC, 65 A
3.2.3 Energy Storage
Another approach that is used to validate the model is to compare the rate of
discharge proportionality factor. The value according to [17] is of 0.283 %SOC/Ah. The
energy content of the system is therefore evaluated in terms of Ah.
For an input average voltage of 47.2 V, and a discharge power of 3.3 kW, there
is a CUITent of 69.9 A. Thus, for 3 hours of operations, 209 Ah of energy is expected. This
is verified by discharging the battery at 69.9 A for 3 hours. Figure 3.5 below shows the
SOC drop from 80% to 23 %.
- 30-
80
70
60
?i
Ü 50
0
(/)
40
30
20
0 0.5 1.5 2 2.5
lime [hl
Fig. 3.5: 3.3 kW VRB discharge at 69.9 A for 3 hours (209 Ah)
The rate of discharge is thus:
Rate of Discharge = 0.57 *100 = 0.272 %SOC / Ah (17)

209
The system energy content is therefore close to 209 Ah, as the VRB IS
discharged from 80% to 23% in 3 hours.
3.2.4 VRB Model Discussion
The following issues need to be considered in battery modeling: battery

discharge behaviour, rate-dependent capacity [21], temperature effects and capacity
fading.
The developed battery model does a good preliminary job at capturing the
discharge behaviour and temperature effects of a VRB. Rate-dependence on the other
hand is Iess of an issue when dealing with flow batteries, as the role of the flow pump is
to ensure that the electrode surface is in contact with the maximum concentration of
- 31 -
active species during charge and discharge. Capacity fading, which is mainly an issue for
Lithium-Ion batteries, is not mode1ed either.
This VRB model has several advantages. It accounts for a large number of
battery operating losses in a simple manner which makes it computationally efficient yet
accurate, and thus well suited for short-term as well as long-term simulations. This also
allows for accessibility and scalability. Its computational simplicity also allows its
integration in real-time simulations, which can combine real and simulated elements to
provide experimentally validated results. Furthermore, this mode1 also accounts for
thermal effects, transients and dynamic SOC behaviour.
However, its limitations lie in the fact that chemical reactions and life-cycle
issues are not represented. Furthermore, CUITent limits and saturations are not mode1ed. It
is thus left to the user to take into account what is reasonable in terms of maximum power
operation, as well as battery life-time.
In this research project, we are interested in studying the battery's behavior and
efficiency within a wind system. The developed model appears to meet these
requirements. However additional experimental data is required in order to fully validate
the mode!.
This is not a major issue since the battery software implementation can be easily
updated to reflect any changes to the model at a later stage. What is more important is the
methodology applied in this project to determine, integrate and analyze the battery mode!.
That approach, which is covered in this report, can then be repeated for different types of
batteries and models, as they become available.
3.3 Li-Ion Model Validation
3.3.1 Cell Voltage
Figure 3.6 shows the individual internaI cell voltage against the SOC, which
varies between 2.7 V and 4.2 V. The curve matches the Li-Ion data found in [27], [30]
and [31].
- 32-
4.2,------,----,---,--,---,---,---.,--,--------,----,.,
3.8
Q)
~ 3.6
g
~ 3.4
~
~ 3.2
-=
3
2.8
o 10 20 30 40 50 60 70 80 90 100
SOC[%]
Fig. 3.6: Li-Ion battery cell voltage versus SOC
3.3.2 Dynamic Charge and Discharge Cycles
The dynamic behavior is examined by looking at the charge and discharge of a

30 cell 40 Ah rated Li-Ion battery. The Li-Ion battery can be operated between 0% and
100% SOC (also referred to as discharge capacity).
3.3.2.1 Full Discharge Cycle
Figure 3.7 shows a Li-Ion full discharge cycle, at a constant CUITent of 40 A.

The voltage and battery output power can be observed as th~ SOC varies between 100%
and 0%. The results agree with Li-Ion data found in [27], [30] and [31].
3.3.2.2 Full Charge Cycle
Figure 3.8 shows a Li-Ion full charge cycle, at a constant current of 40A. The
voltage and battery output power can be observed as the SOC varies between 0% and
100%. The results agree with Li-Ion data found in [27], [30] and [31].
- 33 -
cft.100~
ü 50
oCf)
OL-__L -_ _ ~ _ _~_ _-L__~__~____L -_ _~_ _~==~
o 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

lime [hl
i::r • • • • : : : , ,3
ë
:>
0 0.1 0.2 0.3 0.4 0.5
lime [hl
0.6 0.7 0.8 0.9
i::t'--------------:~.'~:-------,-------:~J
~ 0 0.1 0.2 0.3 0.4 0.5
lime [hl
0.6 0.7 0.8 0.9 1
Fig. 3.7: 40 Ah Li-Ion battery discharge at 40 A for 1 hour (100% initial SOC)
cft. 100
ü 50
o(J)
O~=-~--~---L--~----L---~--~---L--~--~
o 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
lime [hl
i~F,
Ë 0 0.1
: : , : : : ; :J
0.2 0.3 0.4 0.5
lime [hl
0.6 0.7 0.8 0.9 1
Q2 Q3 Q4 Q5 Q6 0.7 0.8 0.9

lime [hl
Fig. 3.8: 40 Ah Li-Ion battery charge at 40 A for 1 hour (0% initial SOC)
- 34-
3.4 Standardized Profile for Experimental Tests
A test system should be used for model validation, battery performance

assessment and technology comparison. It should allow for voltage and CUITent
measurements in time and frequency domain. Tests should include effects oftemperature
on the model, and therefore sorne type of climate control around the battery would be
required. The test equipment should allow readings of the SOC and available energy.
Sorne tests will also require fast changes in the load, and fast switching between charge
and discharge, and therefore the appropriate controls are needed. A series of possible tests
for battery validation was indicated in section 1.3.4.
However, there are currently no published standards for testing battery models
III wind applications. A standardized test profile would provide a methodological
approach to validating battery models. Furthermore, it could be used to evaluate battery
performance within a wind system, as weIl as to compare various battery technologies. A
typical charge-discharge profile will thus be developed in this section.
Wind characteristics need to be examined in order to determine how they impact

wind turbine operation. This will allow the development of a typical wind profile that
covers aIl basic operating conditions related to storage in wind energy applications. It will
be used subsequently to provide a possible charge-discharge profile to test a 3.3kW
lndustrial VRB.
3.4.1 Typical Wind Profile
Wind is characterized by availability, speed (instantaneous and mean), direction,

density, turbulence, shear as weIl as gust. The geographicallocation is also important, as
wind behaviour depends on the local geographical setting (e.g.: plains, mountains, etc.).
In this case, the wind direction is of little interest since the wind data is considered in a
single dominant direction. Similarly, the wind shear, which corresponds to the change in
wind speed based on height, will not play a role in a typical wind profile, since the wind
data is examined at a fixed height.
- 35 -
Wind gust corresponds to the sudden changes in wind speed with respect to the
average wind speed. Thus a highly turbulent wind will have higher gusts, at a higher
frequency. In a wind-storage system, the battery operation is mainly affected by the wind
speed and its variations. However, very high frequency variations will not affect the
turbine response, as they are typically filtered out through the turbine' s mechanical
inertia.
In defining a tentative "typical wind profile", wind data will cover the following
situations:
• High mean wind speed: the mean wind speed will be considerably higher than the
rated turbine wind speed. In this case, the battery is expected to mostly charge and
store the excess power which would be otherwise lost.
• Low mean wind speed: the mean wind speed will be considerably lower than the
rated turbine speed, but above the cut-in speed which will allow the turbine to stay
operational. In this case, the battery is expected to mostly discharge and
compensate for the lack ofpower.
• High turbulence: the wind will be highly turbulent and the speed will vary
considerably. In this case, the battery will have to often switch between charge
and discharge operation. This evaluates its capability to respond quickly.
• Geographicallocation: we also need to look at typical wind profiles obtained from

different geographical classifications. In this example, we consider two
landscapes: "Rolling hills, bushes and small trees" as weIl as "flat landscape, open
fields and meadows".
Based on these criteria, the following wind profile is defined, as shown in figure
3.9. The profile is based on real wind data obtained from the Database on Wind
Characteristics, developed by the RISO national laboratories. The data was obtained at
two locations: San Gorgonio in Califomia, U.S. and Tjare in Denmark. San Gorgonio is
classified as a Hill/Scrub site, at a 539 m altitude, while Tjare is classified as a
flat/pastoral site, at an altitude of 4 m.
- 36-
Typical Typical
Low High High
.
Flat! Hillsl
Pastoral Scrubs Mean Mean Turbulence
30
... • ... • ... • ... • ...
1
25 - - - - - - - - - - L - - - - - - - - - - J - - - - - - - - - - -:-
1
1
20 ----------~----------
~ :
.s :
-g 1
1
Q) 15
c-
C/)
"C
C
:s: 10
o~--------~--~~~~--------~--------~------~
o 0.5 1.5 2 2.5
Time [hl
Fig. 3.9: Typical wind profile for storage device
3.4.2 Charge-Dis charge Profile
In a wind-storage system, the battery acts to smooth out the output power of the
wind turbine by charging and discharging accordingly. In order to provide test data for
batteries in wind energy systems, we propose a typical charge-discharge profile, based on
the typical wind profile presented in figure 3.9. The power output profile is obtained
based on a simple DFIG model for a 2 MW (maximum power) turbine, and a simple
control scheme. Since the power output varies between 0 and 2 MW, if we want the
output power to be constant at 1 MW, the battery needs to be able to provide or absorb 1
MW power.
- 37 -
These values are aIl scaled down accordingly for experimental testing using a
3.3 kW VRB. The following charge-discharge profile is thus obtained, as shown in figure
3.10. The battery profile covers 2.5 hours, at a 2 second sample rate.
3000
2000
~ 1000
~
a..
"5 0
a.
"5
o
~ -1000
>
-2000 - - - - - - - - -
1
1
-3000 _________ J. 1_________ _
1 1
1 1
o 0.5 1.5 2 2.5

Tlme [hl
Fig. 3.10: Typical VRB output power profile
The following steps summarize battery model experimental validation using

standardized wind profiles such as the one proposed in this section:
• Apply the standard wind profile to wind generator model and develop a storage
power profile, based upon the operating algorithm.
• Apply the profile to both the battery and the model, acquiring SOC, current, and
other relevant parameters.
• Compare the experimental and model results and determine whether the
realization is representative or not, based on some predefined specifications.
• If revision is required, modify the model parameters accordingly.
- 38 -
3.5 VRB Performance
The VRB performance is obtained in this section using the developed model, for
a 3.3 kW 3 hour system.
3.5.1 VI Transfer Characteristic
The 3.3 kW VRB VI transfer characteristic is shown in figure 3.11 for various
SOC. The battery rated CUITent is 40 A. The battery can potentially be operated for short
periods of time at a maximum CUITent of 80 A.
_ Charge Discharge ""-

"""'E------=----------;:>~ Idle Battery
... ...
<:E----------"'---"..
70~-~-~-~--~-,--~-~--~-~--.
Rated Peak
Discharge Discharge
60 Current Current
SOC
50 80%
Peak 1 Rated
50%
Charge Charge
20%
Current 1 Current
è. 40
<Il
0>
!!!
g 30
20
10
OL-_+-_-L_~ __ L-_~_-L_~ _ _~_~_~
-1 00 -80 -60 -40 -20 o 20 40 60 80 100

Current [A)
Fig. 3.11: 3.3 kW VRB VI characteristic
3.5.2 System Efficiency
The efficiency will depend on both the SOC as well as the input CUITent.
Information regarding battery efficiency under various operating conditions can be used
to determine efficient operating control strategies. Both charging and discharging
efficiencies are discussed.
- 39-
3.5.2.1 Charge
The system de efficiency during charging operation is defined as:
Plnput
17charge = p- (18)
Stored
The efficiency of the battery is obtained for different input currents and SOC, as
shown in figure 3.12 below. The most efficient region is between 35 and 45 A, and varies
slightly depending on the soc. This is the region in which the battery should ideally be
operated. As the current increases or decreases beyond this region, there is a notable drop
in efficiency.
100
95
Nominal
90 Current
Typical
Maximum SOC
85
80%
50%
80 20%
'#.
>-
u
c:
(J)
75
·u
:;::
û:i 70
65
60
55 Typical
Minimum
50
0 10 20 30 40 50 60 70 80
Charge Current [A]
Fig. 3.12: 3.3 kW VRB charge efficiency versus input currents
- 40-
100
95
90 Nominal Current Typical

85 r:;~==F=::::::::::::::::::::--..-.:M:aXimum
JJIl SOC
80%
;? 80 50%
e..... 20%
;.,
()
c:
al
75
ë3
in 70
65
60
55
50
0 10 20 30 40 50 60 70 80
Discharge Current [A]
Fig. 3.13: 3.3 kW VRB discharge efficiency versns input currents
3.5.2.2 Discharge
The battery dc discharging efficiency is then considered, which is defined as

follows:
PElectrochemistry
17disch arg e = p (19)
Output
Similarly to 3.5.2.1, the efficiency of the system is considered for various SOC
and currents. The result is shown in figure 3.13. The most efficient region in this case is
also identified to be between 35 and 45 A. Similarly, the system efficiency drops for
higher and lower currents.
3.5.2.3 Efficiency for Various Operating Conditions
Tables 4 and 5 below present the various efficiencies of a 3.3 kW VRB for its
full range of SOC and power conditions, for both charge and discharge cases. The charge
and discharge powers are provided in p.u. Note that this refers to the power at the output
- 41 -
of the battery, rather than at the cell stack. The efficiency varies similarly in both charge
and discharge cases, with respect to the SOC and power.
Table 4: Charge Efficiency for Various Operating Conditions
P (p.u.) 2 1.5 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1
P(W) 6600 4950 3300 2970 2640 2310 1980 1650 1320 990 660 330
SOC
0.95 0.829 0.647 0.861 0.862 0.861 0.859 0.854 0.644 0.826 0.79 0.712 0.46
0.9 0.825 0.644 0.859 0.861 0.861 0.859 0.855 0.847 0.83 0.79 0.723 0.485
0.85 0.822 0.842 0.858 0.86 0.861 0.86 0.856 0.848 0.832 0.801 0.73 0.5
0.8 0.819 0.84 0.857 0.859 0.86 0.859 0.856 0.849 0.834 0.803 0.735 0.51
0.75 0.817 0.838 0.856 0.858 0.859 0.859 0.856 0.849 0.835 0.805 0.738 0.519
0.7 0.815 0.836 0.855 0.857 0.858 0.858 0.856 0.849 0.835 0.807 0.741 0.527
0.65 0.812 0.834 0.854 0.856 0.857 0.857 0.855 0.849 0.836 0.808 0.744 0.533
0.6 0.81 0.833 0.852 0.855 0.856 0.857 0.854 0.849 0.836 0.809 0.746 0.539
0.55 0.808 0.83 0.85 0.853 0.855 0.855 0.854 0.848 0.835 0.809 0.748 0.544
0.5 0.805 0.828 0.849 0.851 0.853 0.854 0.852 0.847 0.835 0.81 0.749 0.549
0.45 0.802 0.826 0.846 0.849 0.852 0.852 0.851 0.846 0.834 0.809 0.75 0.554
0.4 0.799 0.822 0.844 0.847 0.849 0.85 0.849 0.644 0.833 0.809 0.751 0.559
0.35 0.795 0.819 0.841 0.844 0.846 0.847 0.846 0.842 0.831 0.808 0.752 0.563
0.3 0.79 0.814 0.836 0.84 0.842 0.844 0.843 0.839 0.829 0.806 0.751 0.567
0.25 0.784 0.808 0.831 0.834 0.837 0.838 0.838 0.834 0.825 0.803 0.75 0.571
0.2 0.776 0.8 0.823 0.826 0.829 0.831 0.831 0.827 0.818 0.797 0.746 0.573
0.15 0.763 0.787 0.81 0.813 0.816 0.818 0.819 0.816 0.807 0.788 0.739 0.574
0.1 0.739 0.763 0.786 0.79 0.793 0.795 0.795 0.793 0.785 0.767 0.722 0.568
0.05 0.681 0.702 0.723 0.727 0.73 0.732 0.733 0.731 0.725 0.709 0.671 0.538
Table 5: Discharge Efficiency for Various Operating Conditions
P (p.u.) -2 -1.5 -1 -0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1
P(W) -6600 -4950 -3300 -2970 -2640 -2310 -1980 -1650 -1320 -990 -660 -330
SOC
0.95 0.732 0.801 0.847 0.853 0.857 0.86 0.86 0.855 0.844 0.821 0.771 0.644
0.9 0.714 0.791 0.843 0.85 0.855 0.859 0.86 0.856 0.847 0.825 0.778 0.654
0.85 0.7 0.784 0.84 0.847 0.853 0.858 0.859 0.857 0.848 0.827 0.782 0.661
0.8 0.689 0.778 0.837 0.845 0.852 0.856 0.858 0.857 0.848 0.829 0.784 0.665
0.75 0.678 0.773 0.834 0.843 0.85 0.855 0.857 0.856 0.849 0.83 0.786 0.669
0.7 0.668 0.768 0.832 0.841 0.848 0.854 0.856 0.856 0.849 0.83 0.788 0.672
0.65 0.657 0.763 0.829 0.838 0.846 0.852 0.855 0.855 0.848 0.831 0.789 0.674
0.6 0.646 0.758 0.826 0.836 0.644 0.85 0.854 0.854 0.848 0.831 0.79 0.677
0.55 0.635 0.752 0.823 0.833 0.842 0.848 0.852 0.853 0.847 0.831 0.791 0.678
0.5 0.622 0.746 0.819 0.83 0.839 0.846 0.85 0.851 0.846 0.83 0.791 0.68
0.45 0.607 0.74 0.816 0.826 0.836 0.843 0.848 0.849 0.844 0.829 0.791 0.681
0.4 0.588 0.733 0.811 0.822 0.832 0.84 0.845 0.847 0.842 0.828 0.79 0.682
0.35 0.564 0.724 0.805 0.817 0.827 0.836 0.841 0.843 0.84 0.826 0.789 0.683
0.3 0.523 0.713 0.798 0.811 0.821 0.83 0.836 0.839 0.836 0.823 0.787 0.683
0.25 0 0.699 0.789 0.802 0.813 0.823 0.829 0.832 0.83 0.818 0.784 0.681
0.2 0 0.68 0.776 0.79 0.802 0.812 0.819 0.823 0.821 0.81 0.777 0.678
0.15 0 0.65 0.755 0.77 0.783 0.794 0.802 0.807 0.806 0.797 0.766 0.671
0.1 0 0.59 0.715 0.732 0.746 0.758 0.768 0.775 0.776 0.768 0.741 0.652
0.05 0 0 0.598 0.62 0.638 0.654 0.668 0.677 0.682 0.68 0.66 0.587
These results are useful when developing operating strategies for wind energy
storage systems. The matrices need to be taken into account, and the operating strategy
optimized to make sure that charge and discharge operations are done at the most efficient
operating points when possible. This data also needs to be taken into account when sizing
the battery.
- 42-
Chapter 3: Madel Validation and Performance
3.5.3 Dynamic Response
In order to assess the battery dynamic performance, the worse case transition is
considered: the operation is switched from a charging CUITent of 80 A to a discharging
CUITent of -80 A. Figure 3.14 shows that the battery output voltage takes 52 ms to reach
steady state when switching from charge to discharge. It takes 47 ms to reach steady state
when switching from discharge to charge. The transition time is not dependent on the
SOC.
The impact of the e1ectrode capacitance is noticeable when observing the battery
voltage. The transients could thus impact the battery operation at variation rates of 19 Hz
and above. However, the mechanical inertia in a DFIG wind system will typically play a
major role in damping the impact of rapid wind fluctuations, by acting as a low pass filter.
60
10%
58
56 Max Charge to
Max Discharge
54 Response lime
5>
'Q)' 52
Cl
;g
ë5 50
> -<Max Discharge to
>1
~
:ê<Il 48
Max Charge
III 1
Response lime
46
1
44
190 %
1
42
1 1
40
0.5 0.52 0.54 0.56 0.58 0.6 0.62 0.64 0.66
lime [s1
Fig. 3.14: Worse case charge to discharge transition (80 A currents)
- 43 -
Idle Current
150,---,----,---,----,---~--_,----,_--~--_,--~
2x Rated Rated Charge Rated
Current Current 2x Rated
Disharge
=: Current
Current
[ SOC
1-~~r==t==S~ 80%
-l50%
100 20%
1 1
>
ID
Cl 1 1
~
(5
> 1 1
50
1 1
1 1
1 1
OL___ ~ ___ L_ _ ~_ _ _ _L __ _~_ _~_ _~L_ __ ~ __ _L--~
-100 -80 -60 -40 -20 0 20 40 60 80 100

-c::r----C-h-a-rg-e------:l:>~ Current [A] ...<r----D-is-c-ha-r-ge-------;:>~
...
Fig. 3.15: Li-Ion VI characteristic
3.6 Lithium-Ion Performance
The Li-Ion battery performance is obtained in this section using the developed
model.
3.6.1 VI Transfer Characteristic
The 40Ah Li-Ion battery VI transfer characteristic is shown in figure 3.15 for
various soc.
3.6.2 System Efficiency
In this section, both charging and discharging efficiencies are obtained, as a

function of current and soc.
3.6.2.1 Charge
The system dc efficiency during charge is defined as:
- 44-
P1npu/
17charge =~ (20)
Stored
The efficiency curves are shown in figure 3.16. The battery efficiency is found
to be 97% for the rated CUITent. It drops linearly as the CUITent increases.
3.6.2.2 Discharge
The system dc efficiency during charge is defined as:
PElectrochemistry
17disch arg e = p (21)
Output
The efficiency curves are shown in figure 3.17. The efficiency for discharge is
similar to the efficiency during charge, increasing with the CUITent amplitude. It is also
found to be 97% at the rated CUITent.
Rated Charge Current

100
98
96
SOC
94 80%
--- ~-.J 50%
92 20%
~
~
>.
u
cQ) 90
~
W 88
86
84
82
80
0 10 20 30 40 50 60 70 80
Charge Current [A]
Fig. 3.16: Li-Ion battery charge efficiency versus current
- 45 -
Rated Discharge Current

100
98
96
94
SOC
80%
50%
~ 92 20%
e.....
>-
u
c:::
al
90
·ü
:;::
û:i 88
86
84
82
80
0 10 20 30 40 50 60 70 80
Fig. 3.17: Li-Ion battery discharge efficiency versus current
122
1
1 1
120 10%
1 Full Charge to 1 1
118 1 Full Discharge

Response lime 1 1
116 ~I ~I
~
al
Cl
Full Discharge to
.l!! 114 1
Full Charge 1
ë5 Response lime
>
~ 112 1
2 1
êiî
!Il 110 1 1
108 1 1 1
106 1 1 1
90%1 l
104
1 1
0.48 0.5 0.52 0.54 0.56 0.58 0.6 0.62 0.64
lime [s]
Fig. 3.18: Worse case charge to discharge transition (Initially 50% SOC, 80 A currents)
- 46-
3.6.3 Dynamic Response
The worse case transition is considered: the operation is switched from a

charging current of 80 A to a discharging current of -80 A. Figure 3.18 shows that the
battery output voltage takes 19 ms to reach steady state when switching from charge to
discharge. It also takes 19 ms to reach steady state when switching from discharge to
charge. The dynamic response is not dependent on the SOC.
3.7 Li-Ion and VRB Performance Comparison
The battery models were validated and the battery performance was obtained,
inc1uding the VI transfer characteristic, the efficiency and the dynamic response. In terms
oftechnical performance, the Li-Ion battery demonstrated better results, inc1uding:
• Higher efficiency: 97% Li-Ion battery efficiency compared to 86% VRB

efficiency during charge and discharge at the rated currents.
• Faster dynamic response: 19 ms response time for the Li-Ion battery compared to
54 ms for the VRB.
• SOC operating region: Similar to the VRB output voltage, the Li-Ion battery
output voltage is linear for a SOC range of 20% to 80%. Both batteries should be
operated in this range in wind energy applications, to avoid overcharge or
undercharge issues.
The VRB advantages over the Li-Ion inc1ude better scalability, longer lifetime,
and better economical performance potential (inc1uding reduced cost per kWh for larger
systems). The increased performance of the Li-Ion battery might justify its use in certain
situations where very high efficiency is needed, however the VRB appears to be better
suited for large scale applications.
- 47-
3.8 Conclusions
VRB and Li-Ion battery models were developed and implemented in

SIMULINK. A systematic approach towards battery rnodeling and validation was
presented. The batteries performances were studied and the VI transfer characteristics as
well as transient responses were determined. Furthermore, data regarding the operating
efficiency under various conditions was obtained. This provided insight on how to
optirnize the operation of these batteries, as well as a basis for cornparing the two battery
technologies.
The next step will consist of looking at the design of wind-energy storage
systems. The VRB, which so far seems better suited for large-scale applications, will be
integrated into large wind energy DFIG systems. VRB storage design and integration for
DFIG systems will be examined using the developed model. Model scalability will also
be considered.
- 48-
Chapter 4: Wind Energy Storage System Design
Chapter 4
Wind Energy Storage system design
4.1 Introduction
A VRB model has been developed and validated. It was subsequently used to
assess the battery performance, induding transient operation, transfer characteristic and
efficiency. The battery integration in a wind energy system will now be discussed. There
are two main issues that need to be considered in wind-storage design, which are the
battery sizing and the battery interface.
Battery sizing issues will first be considered. The mode! scalability for
integration with a large wind turbine (> 1 MW) is then discussed. Finally a dc-dc chopper
interface is developed to control the flow of the current in and out of the battery. Its
impact on battery performance is also considered.
4.2 Battery Sizing
Battery sizing for wind energy applications is a major issue, because it directly
affects the economics of the project. A weIl sized battery can be the difference between a
feasible and non-feasible project. The advantage of having an accurate battery model is
that it can be used for system testing in specific applications over a desired timescale.
This allows the validation of the battery system design pararneters. When designing a
VRB system, there are 3 sizing pararneters that need to be considered: the battery rated
power, the battery maximum power and the battery energy capacity.
The battery sizing will be dependent on the operating needs of the battery. A
battery will play two roles in power systems: output power smoothing (short-term) and
load leveling (medium and long-term). Typically, other flow batteries such as NaS will
have a power rating of 80% of the wind farrn capacity they are associated with.
The battery cell rating, as defined by the manufacturer, is the power at which the
battery cell will be operated at its most efficient point. However, under contingencies, the
- 49-
battery needs to be operated at higher than normal power for a short period of time
(seconds), defining the maximum battery power. The energy capacity on the other hand
will depend on the maximum amount of energy stored. The following are sorne of the
issues that need to be taken into consideration in wind energy battery sizing:
• The battery system should play a role in both load leveling and power smoothing
applications, in order to maximize the system benefits.
• The battery system must be able to displace the needed energy from low-demand
periods to high-demand periods.
• There is a large capital cost associated with a VRB, however the cost per kWh
decreases as the energy storage capacity increases;
• The battery needs to be optimized so as to reduce size and cost. On the other hand,
larger capacity allows more day-ahead purchases of balancing power, which in
return avoids expensive spot purchases.
• The battery system must be able to provide, at any given time, enough energy to
allow the startup of backup generators such as diesel. Thus, operation at above
rated currents in sorne transition cases must be possible for short periods of time.
• Typically, VRB size varies between 15% and 25% of the wind farm rated
capacity.
In particular, in remote wind-diesel systems, the system should be optimized so

as to reduce emissions and operating costs related to diesel generation. In other words, the
diesel system should be shut down when possible. The battery however must also ensure
that the diesel system has enough start-up time when the wind resource becomes scarce,
without any electrical interruption to the local community.
Diesel system operation should be avoided below 45% of its capacity, where it
becomes very inefficient. The wind-battery system must be designed, along with the
control algorithms, to either allow the diesel generator to run at near-full capacity, or
allow it to shut down.
- 50 -
Based on typical sizing in recent projects [17], the integration of a 250 kW VRB
system will be examined for a wind turbine in the 1 MW range.
4.3 Scaling the VRB Model
4.3.1 System Requirements
The VRB will be integrated on the dc bus of a 1 MW DFIG wind turbine. Since
the 3.3 kW model efficiency and output curves agree with results found for a 42 kW VRB
[17], it will be scaled up accordingly. The VRB will have a nominal output voltage of 120
V, and thus a 100 celI stack will be needed. A system of six 42 kW VRB batteries will be
used, for a total power of 0.252 MW (0.504 MW peak).
4.3.2 Modeling a 42 kW Industrial VRB
The 42 kW VRB model is derived using the same method covered in chapter 2.
Similarly, the calculations are based on the worse case operating point at the end of the
discharge cycle. The losses are divided into 15% internaI and 6% parasitic. At that point,
a minimum voltage of 105 V is reached. The maximum discharge current is 400 A. Thus
for the battery to be able to provide 42 kW with 21% losses, the celI stack output power
should be:
42000
P,·tack = 1-0.21 = 53164.55 W (22)
4.3.2.1 Rinternal and Rparasitic
Rintemal accounts for 15% of the losses, which corresponds to 7974.7 W. At an

operating current of 400 A, RintemaJ is found to be 0.0498 Q. The parasitic los ses are
separated into fixed and variable losses, as folIows:
PparaSitiC = P flXed + k(~~~) =1063.3 + 2126.6(~~~)( :000) (23)
The parasitic fixed and variable losses are obtained as follows:
- 51 -
1052 )
R flXed = =10.37 n (24)
( 1063.3
106.3 3( SOC
Istack )
= 1.0126( Istack ) (25)

105 SOC
4.3.2.2 Transient Parameters
Rintemal can now be divided into Rreaction (0.03 n) and Rresistive (0.0197 n), using
the same proportions as in section 2.3.5. Ceiectrodes is determined to be 0.06 F for a 100 cell
stack, where each cell has a 6 F capacitance.
Discharge Charge
....c: :> Idle Battery -< :>
180
SOC
80%
160 50%
Full 20%
140
Charge
Current
120
> 100
Q)
0>
$
(5 80
>
60
40
20
O L -_ _ _ _L -_ _ _ _L -_ _ _ _L -_ _ _ _L -_ _ _ _L -_ _ _ _L -_ _ _ _L -_ _~
-400 -300 -200 -100 o 100 200 300 400

Current [A]
Fig. 4.1: 42 kW VRB model VI characteristic (Various SOC)
- 52-
4.3.2.3 Model Implementation
The new values for Rreaction, Rresistive, Celectrodes and RPixed are integrated into the
initial SIMULINK VRB mode!. We also modify Ipump , the number of cells as weIl as the
time and power ratings. Figure 4.1 shows the 42 kW battery transfer characteristic for
various SOC values.
4.3.3 VRB Thevenin Equivalent
The battery model steady state Thevenin equivalent can be obtained for various
SOC. The RThevenin value corresponds to the slope of the VI transfer characteristic, while
the V Thevenin value can be obtained from the open circuit voltage, èorresponding to idle
battery operation. Table 6 shows the Thevenin equivalent parameters, for a SOC of 20%,
50% and 80%.
Table 6: Battery Steady State Thevenin Equivalent
SOC RThevenin Vrhevenin
20% 0.05 132 V
50% 0.05 139V
80% 0.05 146 V
4.3.4 Multiple Stacks
A 42 kW battery is not sufficient for a 1 MW DFIG, where there might be a

need to inject or extract up to 500 kW at any time. One possible solution which has been
tested previously is to use six 42 kW VRB stacks in series [17] in order to build a 252 kW
system.
In terms ofmodelling, there are two options: scaling up one model to 252 kW or
using several 42 kW models in series. The second alternative offers better insight
because, typically, separate stacks can have varying efficiency (up to 3% [17]). Having
- 53 -
several stack models allows the consideration of these various efficiencies. Figure 4.2
below shows an example of two stacks in series.
+
+ V Stack
CElectrodes 1
RFixed Losses Ipump

-+ Losses: RReaction 1
1
1
1
1
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ J1
Total Electrolyte
Energy Storage
RResistive Losses 1
(SOC)
+ ---------------- -----------
1
1
1
VStack
1
1
1
è:': CElectrodes 2
1
~ 1 Ipump
co : RFixed Losses
al : -+ Losses
> : 1
1
1L ___________________________ J1
Fig. 4.2: Two VRB models in series
The varying efficiency of each stack can be considered by deriving separate

parameter values for the parasitic fixed losses (Rfixed losses), the reaction losses (Rreaction) as
weIl as the resistive losses (Rresistive losses). The associated pump losses (Ipump) will pro duce
the same parasitic CUITent in each sub-model; however the resulting parasitic losses will
vary due to the difference in voltage across the parasitic branch. The internaI stack
voltages of each stack will be the same.
- 54-
Chapter4: Wind Energy Storage System Design
While there may be several physical pumps and tanks in such a system, the
developed model represents these components in the following way:
• Stack modelling: The developed battery model is primarily a stack model, which
represents the stack voltage and the associated operating losses. The voltage of
each stack is dependent on the SOC, while the operating losses are modelled
through various circuit parameters. A stack model is used for every battery.
• Tank modelling: The tank physical characteristics are not modelled in detail,
because they do not have a major impact on the battery operation. It is assumed
that the tanks are optimized to favour a satisfactory electrolyte flow to the stacks.
The tanks are simply represented by a common variable which represents the
overall SOC, shared by the various stack models.
• Pump losses: The pump losses are directly dependent on the operating conditions
(SOC, current amplitude and battery voltage). For higher power flow, the pump
losses will be greater. While physically, the pumps are associated with the tanks,
each stack sub-model will represent its associated operating pump losses.
4.3.5 252 kW VRB System
The 252 kW VRB model system in shown in figure 4.3 below. Six sub-models
are used. Individual stack efficiency can vary by up to 3%. This can be taken into account
by modifying the individual sub-model parameters (Rfixed lasses Rreaction, Rres lasses)
accordingly .
. This battery bank can be operated at a maximum of 400 A. With a 1200 V dc

bus, this would translate into an output power of 480 kW, which meets our needs. In
terms of energy content, the VRB tanks should be sized based on economic factors, as
weIl as the requirement to operate the battery between 20% and 80% SOC at aH times.
- 55 -
VRB Battery System

252kW-200A
J
:::r::
Staek Model
42KW -100 Cells =
=
-,-
1
:::r::
SOC
Staek Model
42KW-100 Cells =
=
Battery Tanks 1
:::r::
Staek Model
=
&
=T
~r P Storage-
42KW - 100 Cells
1
:::r:: ....
nnnnr'\
=
~~
Staek Model
42KW-100 Cells =
1
:::r::
Staek Model
42KW-100Cells
=
=
2-quadrant
de-de chopper
1
:::r::
Staek Model
42KW-100 Cells =
=
-,-
1
Fig. 4.3: 252 kW battery bank
4.4 Battery Converter Interface
4.4.1 Buck-Boost dc-dc Converter
A dc chopper interface is used to control the CUITent direction and amplitude

through the battery. Figure 4.4 below shows the implementation of such a dc chopper in
SIMULINK using the VRB model. The DFlG bus has been initially modeled as an ideal
dc source. The dc chopper efficiency and controls will be examined for a 42 kW battery
model.
4.4.2 Inductance Design
Part of the converter design involves determining an appropriate value for the
battery side inductance. There will be a dc current ripple which will be dependent on the
value of that inductance. This ripple will result in battery heating, as weIl as increased
battery cycling, which can have a negative impact on the battery life.
- 56-
Buck-Boost DClDC Converter:

Dis::rete
I-re' 14---1
, - - - - - - - - - - I mod
Universal Bridge
1 ann
!.l
T
-L
l
'-------~IDC
Discrete.
Ts = 0.001 s.
Fig. 4.4: Dc chopper implementation for VRB
A larger inductor value will result in a smaller current slope, and thus a smaller
ripple. On the other hand, this results in a smaller response time for the controller. While
the ripple can not be fully eliminated, a ripple of 10% or less of the dc CUITent magnitude
is deemed acceptable.
The dc bus voltage is 1200 V, and a CUITent ripple of 10% or less for currents of
up to 400 A is required. Thus the value of the inductance is obtained the following way:
Switching Frequency = 1080Hz (26)
(27)
L = I1Vxl1t = 1200 x 4.62e-4 - 14 mH (28)

l1i 40
- 57 -
An inductor of 14 mH provides the best trade-off between response speed and

CUITent ripple. It produces a ripple of 10% (figure 4.7), while generating an acceptable
response time (figures 5.5 and 5.12).
4.4.3 Converter Controls
Figure 4.5 below shows that, by setting a reference current indirectly through
the modulation index of the dc chopper, we are able to charge and discharge our battery at
400 A. The battery model used in this specific example is rated at 42 kW - 1 minute.
f: o 20 40 60
lime [51
80 100 120
L~r : :
> 0 20 40
1.
60
lime [51
, ":
80 100
·
120
è..
!::f : : t : ' J
> 0 20 40 60
lime [51
80 100 120
Fig. 4.5: Charging and discharging the 42 kW VRB using a dc chopper

(400 A and -400 A respective current rates, 160 V, 1 min-42 kW VRB)
Since the current rate is dependent both on the modulation index, as well as the
battery output voltage which varies with the SOC, a PI control scheme is used as shown
in figure 4.6 below.
- 58 -
1 Reference
----I~~
0 I:l
~U - - - - I..~
- - - - - 1..
oc-oc Chopper
Modulation Index
r
1 Measured
Fig. 4.6: Dc-dc chopper PI control
The response time of the controller can be observed to be negligible in figure

4.7, which shows the charge and discharge currents response to the reference. There is a
ripple on the dc current of 30 A. This translates into about 10% for a 300 A CUITent. This
noise ripple depends on several factors including the PI controller settings and sample
time, as weIl as the inductance used between the chopper and the battery.
~ 400
ë 1
~
:; 200
ü 1
CD 0
t)
t:
1
~ -200 1
CD 1
'Iii
0: -400
2 3 4 5 6 7
lime [s1
400
~ 200 1
ë
ê
:J 0
Ü 1
Ü -200
0
-400
2 3 4 5 6 7
lime [s1
Fig. 4.7: 42 kW VRB current control using the dc chopper

(180 V dc bus, varying current reference)
4.4.4 VRB de efficieney
The system dc efficiency during charge and discharge, including chopper losses,
will be examined in this section. As was shown in chapter 2, the battery is more efficient
for higher SOC. The SOC will be arbitrarily fixed at 50% as the reference CUITent is
- 59-
varied in order to determine the system efficiency at various operating points. The charge
and discharge efficiencies are defined as follows:
PDC BUS
1]Discharge =P (29)
Battery _ Internai
- PBattery Internai
1]Charge - p (30)
DC_BUS
The discharge is found to be most efficient at currents between 175 and 215 A,
close to the rated current (200A). The battery system efficiency reaches 82.5 % in this
region. It drops considerably for currents below 100 A and above 400 A. This can be seen
in figure 4.8. In figures 4.9, charging operation is found to be most efficient at currents
between 120 A and 170 A. The battery system efficiency also reaches 82.5% in that
reglOn.
100
90
~
e.....
>- 80
u
c:
Q)
70
~
LU 60
50
-400 -350 -300 -250 -200 -150 -100 -50 0
4
x 10
8
~6
~
a.. 4
(J)
::l
cc 2
g
0
-400 -350 -300 -250 -200 -150 -100 -50 0
Fig. 4.8: Discharge efficiency versus current for VRB-Chopper

160 V de bus, 50% SOC, 42 kW VRB
- 60-
100
90
<ft
>-
u
80
c
Q)
ë3 70
in 60
50
0 50 100 150 200 250 300 350 400
Charge Current [A]
4
x 10
0
~
~ -2
0
~
c.. -4
CI)
::J
en -6
ü
Cl
-8
0 50 100 150 200 250 300 350 400
Charge Current [A]
Fig. 4.9: Charge efficieney versus eurrent for VRB-Chopper

160 V de bus, 50% SOC, 42 kW VRB
The VRB efficiency values agree with the ones presented in [17]. In field trials,
overall dc battery efficiencies were found to be between 82% and 85%.
In comparison with the battery efficiency prior to the dc chopper

implementation, the efficiencies have dropped slightly. The charge efficiency was found
to be around 86% in chapter 3 in the best operating regions. This translates into a drop in
efficiency of 3.5% during charge and discharge. Typically converter losses vary around
5%, and thus the results found are satisfactory. This brings the overall system round-trip
efficiency to 68%.
4.5 Conclusions
In this chapter, a dc-dc chopper interface was developed for the VRB model, in
order to control the power flow in and out from a dc bus. VRB design and sizing issues
were discussed, and a 42 kW battery model was developed for larger wind turbine
- 61 -
applications. Multiple models were used in series to represent a 252 kW multiple-stack

VRB system.
The system performance was evaluated including the transient response and
system efficiency. The VI transfer characteristic was obtained for various SOC, and the
model Thevenin equivalent was derived. The converter losses were found to reduce the
system dc efficiency by 3.5% during charge and discharge. The VRB-chopper system can
now be integrated in various DPlG systems in order to smooth out turbine power output
under fluctuating wind conditions.
- 62-
Chapter 5: Wind Energy Storage System Implementation
Chapter 5
Wind Energy Storage System Implementation
5.1 Introduction
In the previous chapter, a VRB storage system was designed for large DFIG
wind turbines as a solution to storing and releasing energy as required. A dc-dc interface
was deve!oped to control the power flow. Its impact on system performance was also
examined. This system will now be integrated in a DFIG system. The storage system will
be used to level the turbine output power, by charging or discharging accordingly.
This will be done using various available DFIG models: the detailed DFIG
system model, the average DFIG system model and the simple DFIG system mode!.
These models pro vide varying accuracies and simulation speeds.
The detailed mode! is well suited for short-term simulations (seconds). The
average mode! on the other hand is better suited for medium and long term simulations
(minutes and hours), and provides a good tradeoff between accuracy and simulation
speed. Finally, the simple model is best suited for real-time simulations. The first two
cases will adopt a localized storage scheme, while the last case will use a centralized
storage scheme.
5.2 Storage Schemes
There are two possible schemes to consider: local and centralized storage. In the
local storage scheme shown in figures 5.1, the storage system is integrated on the dc bus
of the DFIG system. The advantages include a dedicated storage device for each DFIG
system. There is no need for an additional dc-ac converter to connect the storage system
to the grid, as the ac si de converter of the DFIG system is used for that purpose. The
objective is to smooth the power delivered to the grid by the turbine. This can be
summarized in the following equation:
- 63 -
(31)
POrid constant can be made constant by controlling the power flowing in and out
of the storage device.
The centralized storage scheme is shown in figure 5.2 be1ow. In this setup, the
storage system is connected directly to the grid, independently of the DFIG. In that setup,
the storage can be sized to accommodate severa! turbines simultaneously. However there
is a cost disadvantage because an additional dc-ac inverter is required. This is thus better
suited for larger wind farms, since the cost per kWh of a VRB drops as its capacity
increases.
GRID
DFIG
PConverter
... .
PStator
~~ \ ~~ t
\
\
\
\\
1
...
r
\
PS!orage \
\
\
Controllable
de source r \\
\
2-quadrant
dc-de chopper
Fig. 5.1: Local DFIG storage system
- 64-
Varying Wind
1(\) I--..~~ P Wind

Wind Turbine P Grid
P Storage
~
VRB dcl dcl

Storage
Deviee 1 dc 1 ac
Fig. 5.2: Centralized storage system
5.3 DFIG DetaHed System Model
5.3.1 Mode} Operation without Storage
The integration of a VRB storage system in the DPlG wind turbine will be
examined. Since the battery output is a dc voltage, and the battery is controlled through a
dc-dc chopper, there is an interest in placing the storage system on the dc bus of the
DFIG. Injected currents will affect directly the line-side of the DPlG system without any
negative impact on the DFIG operation.
The DFIG detailed model available III the SimPowerSystems SIMULINK

toolbox is used. The model is adjusted to reflect the operation of a single 1.5 MW wind
turbine. This DPlG system operates best at a rated wind speed of 10 mis.
- 65 -
5.3.2 ControIIing the Battery System
The battery is controlled through the de-de chopper, using a PI block, which sets
the CUITent in and out of the battery current to the desired reference, as discussed in
section 4.4.
The VRB storage system will be used to regulate the line-side output power of
the DFIG turbine. This will result in a smoother power output to the grid as the wind
fluctuates over a short-term (seconds) and medium-term (minutes) time frame.
In order to achieve this, the system controls must be verified. A second PI

controller is used in this case to control the power injected or extracted from the battery.
The error signal is obtained by comparing the mean power output of the DFIG to the
reference power, as shown in figure 5.3 below.
P Storage
Refere.;..;n.:....ce"---_-I~~ 0 __--II.~I;l-I-R-e-fe-r-e-nc-e_l.~
o VRS Storage
System
P Storage
Measured
Fig. 5.3: DFIG system power reference controls
In order to test the controls, the system ability to respond to a varying reference
power is verified in figure 5.4 below. The output power ripple was existent prior to VRB
integration when using the detailed DFIG system model. It can partly be improved by
increasing simulation time step.
The output power does a good job of following the reference power by injecting
or absorbing power using the VRB as required. The response time of the control system is
observed in figure 5.5 below.
- 66-
Q)~
1·~C
~ i
g~
........
Q)~ ; j
:
Q) Q)
0.5
Q5
cr:
~
0
0...
0
1.5 2 2.5 3
:
3.5 4
:
4.5
!
5 5.5 6
:
6.5 7
lime [s1
Ji::~~·~.M 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7

lime [s1
2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7

lime [s1
Fig. 5.4: Power output response (DFIG detaHed model, 252 kW VRB)
50% SOC, 1200 V de Bus, 252 kW VRB
1 1 1 1 1 1 1 1 1
1.4 - - - ~ - - - ~ - - - - ~ - - - ~ - - - -: - - - - ~ - - - ~ - - - -: - - - -:- - - -
1 1 1 1 1 1 1 1 1
1
1
1 1 1 1
1.2 - - - + - -- ----~---~---~---- ---~---~----~---
1 1 1 1 1 1
1 1 1 1 1 1
1 1 1
1
~
1
_. __ .1 __ _ __L ___ L ___ _ _ _ _ L __ _
1 ~
~ 1 1
~
Cl. 0.8 1 1
----1------
______ L ___ ! ___
1 1
~
1
____ L __ _
1
~ 1
1
1
1
1
1
"5 1 1 1
o 0.6 ___ ~ __ _ 1 1 1
T---~----r---T---~----r---
1 1 1
1 1 1 1
1 1 1
1
1 1
0.4 ---+--- ~---+---~----~---~---~----~---
1 1 1 1
1 1
1
1
0.2 '------"_----'._----'-_---"-_--L._---'--_--'---_--'---_-L---.----J
0.9 0.95 1.05 1.1 1.15 1.2 1.25 1.3 1.35 1.4
Time [s]
Fig. 5.5: Power output response time (DFIG detaHed model, 252kW VRB)
50% SOC, 1200 V de Bus, 252 kW VRB
- 67-
The worse case transition response time was tuned down to 0.40 s. This is
satisfactory for this application because, in the case of faster wind variations, the DFIG
turbine will take care of filtering the high frequency wind oscillations.
The next step is to determine the value of the system output reference power
during normal operation. The used DFIG turbine can output 0.85 MWat a rated wind
speed of 10mls, which will be used as the power reference. Thus the storage system will
act to keep the output power at 0.85 MW, regardless ofwind variations.
The storage system reference power will be determined as the difference

between the actual DFIG power output, and the desired one. The overall controls are
shown in figure 5.6 below.
P Grid P Storage
reference_ _--I:~ 0 __ R_ef_er_en_c_e_ _~~ O---j"~G _1_R_ef_er_en_c-je..~ VRS Storage
System
rI"
P Grid
Measured
P Storage
Measured
Fig. 5.6: Ove rail control system
5.3.3 Wind System Operation
ln order to test the validity of the control system, two simple cases are observed.
ln the first case, the wind suddenly drops from its nominal value of 10 mis, to 6 mis.
Figure 5.7 below shows the output power as weIl as mean output power with and without
the VRB storage system on the dc bus.
There is a significant improvement: the power output will remain at an average

of 0.85 MW when the battery storage is active. The injected CUITent on the dc bus, the
drop in the SOC as weIl as the Battery output voltage are shown in figure 5.8.
- 68 -
§'
e.
~
0
a. 0.5
'5
c..
'5
0
0
1 2 3 4 5 6 7
Time [s1
§'
e. ID
Cl
~
ID
> ~
oC(
"0
0.5
ID
iiî
"S
u
(ij
Ü 0
1 2 3 4 5 6 7
Time [s1
Fig. 5.7: Power output after sudden wind drop (DFIG detailed model, 252 kW VRB)
From 10 to 6 mIs, 50% SOC, 252 kW-1 h VRB, 0.85 MW reference power
oC( 800
600
E 400
@
::; 200
Ü 0
Ü -200
0
1 2 3 4 5 6 7
Time [s1
~ 1000~----~------~------~------~------~----~
f::~~~
Jg 1 2 3 4 5 6 7
Time [s1
~:[
~
e.....
ü
0
Cf)
1 2 3 4 5 6 7
Time [s1
Fig. 5.8: VRB system after sudden wind drop (DFIG detailed model, 252 kW VRB)
From 10 to 6 mIs, 50% SOC, 252 kW-1 h VRB, 0.85 MW reference power
- 69-
~
~
~
a... 0.5
"S
c..
"S
o
1.5 2 2.5 3 3.5
lime [s]
1.5 2 2.5 3 3.5

lime [s]
Fig. 5.9: Power output after sudden wind rise (DFIG detailed model, 252 kW VRB)
From 10 to 13 rn/s, 50% SOC, 252 kW-1 h VRB, 0.85 MW reference power
« 200
ë 0
~
::J -200
ü
g 400L-______~______~______~______~________~~
1.5 2 2.5 3 3.5
lime [s]
~ 1000,-------,-------,--------,-------,--------,---,
I::~
Jg 1 1.5 2 2.5 3 3.5
lime [s]
~
:1
0
50 : :
49.99 f :
1 1.5 2 2.5 3 3.5
lime [s]
Fig. 5.10: VRB system after sudden wind rise (DFIG detailed model, 252 kW VRB)
From 10 to 13 rn/s, 50% SOC, 252 kW-1 h VRB, 0.85 MW reference power
- 70-
The next case that is considered is a sudden wind rise from lOto 13 mis. Figure
5.9 shows that the output power is maintained at 0.85 MW when the VRB system is used.
Figure 5.10 shows the operation ofthe VRB system.
In this section, the pre!iminary implementation of a VRB on the dc bus of a

DFIG system was shown. The storage system was used to smooth out the power. This
computationally heavy DFIG system mode! is mostly suited for short-term simulations
(seconds). There is thus a need to examine other possible DFIG system models better
suited for medium and long term simulations.
5.4 DFIG Average System Model
5.4.1 Operation without Storage
Another mode! available through the SimPowerSystem library is the DFIG

system average mode!. This mode! is less computationally heavy since it does not
consider converter switching losses and harmonics. It is ideal for medium-term (minutes)
simulations.
The same wind storage scheme discussed in section 5.3 is used, which consists
of integrating the VRB on the dc bus of the DFIG system, thus taking advantage of the
line-side dc-ac converter.
5.4.2 Controlling the Battery System
In this section, the controls of the storage system are verified. They have been
slightly modified from the detailed model controls. Figure 5.11 shows the response of the
system to a varying output power reference.
Figure 5.12 be!ow shows a worse case transition, where the reference output
power is switched from 0.3 MW to 0.9 MW. The response time is of 0.17 seconds, which
is satisfactory since higher frequency wind fluctuations would be filtered out through the
DFIG mechanical inertia.
- 71 -
§'
è.
.... 1
Q)
s:
0
Cl.. 0.5 1
Q)
u 1
c:
~
-Q)
Q)
a:: 0
1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
lime [51
1
~
è. \
"-
Q)
~ 0.5
\
-
Cl..
::J
C.
l
'S
0
0
1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
lime [51
Fig. 5.11: Power output response (DFIG average model, 252 kW VRB)
0.9
QS ------------- ~-------~-------~-------~------
1 1 1 1
1
1
__ ~ _______ ~ _______ l _______ L _____ _
~ 0.7 - - - - - - - - - - - 1 1 1
~
1 1
1
~
n.
0.6
1 1
---~-----------------------r------
1
1 1
"S
a.
8 0.5 - - - - - - - -- ----~-------~-------~-------~------
1 1
1
1 1
1 1 1
1 1
1 1 1
_____ _______ _______ _______ L _____ _
004 - - - - - - - - ~ ~ ~
1 1
1
1 1
0.3 f----~ - - - - - - -: - - - - - - - -: - - - - - - - ~ - - - - - - - 1- - - - - - -
1 1 1
4.95 5 5.05 5.1 5.15 5.2 5.25

lime [s]
Fig. 5.12: Power output response time (DFIG average model, 252 kW VRB)
(Reference and output power traces)
- 72-
5.4.3 System Operation
In order to test the system, the wind profile shown in figure 5.13 below is used.
Figure 5.14 shows the Average DFIG turbine model power output obtained from this
wind profile, prior to the integration of the VRB storage system.
15,-------,-------,-------,-------,--------.------~
1 1
5 - - -- - - j- - - - - - +- - - - - --t - - - - - - -1- - - - - - -1- - - - - -
1
10 20 30 40 50 60
Time [s]
Fig. 5.13: Wind profile
1
1 1 1 1 1 1 1 1 1 1 1
0.7 - - T" - - -1- - -"'T - - - r - - ,- - - T - ~ -,- - -"1 - - -,- - - -,- - - r --
1 1
0.6 - - t- - - -1- - - -t - - -1- - - -;- - - t- - - -1- - - - t - - -1- - - -1- - - t- - -

1 1 1 1 1 1 1 1 1 1 1
1 1 1
1 1
~ 0.5 _ ..J ___ .l ___ 1_ _ _ -l _ _ _ 1- ___1 _ _ _ .!.... _ _
è.
~
0 0.4
Cl..
'5 1
a. 1 1 1 1 1
'5 0.3 --r------l---r--l---
0 1
1
1
0.2 - - r - - -,- - - ï - - - r - -"1 - - - T - - -,- - - 1 - - - r - - -,- - - r - -
1 1 1 1 1 1 1 l , 1 1
1
0.1 - - +- - - -1- - - -+ - - - 1-- - - --1 - - - +- - - -1- - - -+ - - - 1- - - -1- - - +- - -
0
5 10 15 20 25 30 35 40 45 50 55 60
Time [s]
Fig. 5.14: Turbine output power without storage (DFIG average model, 252 kW VRB)
- 73 -
The turbine output fluctuates considerably, and thus the need for a VRB as a
means to smooth out these variations is justified. Figure 5.15 below shows the results
obtained when the VRB storage system and corresponding controls are added to the
mode!. The power output is centered on the set reference of 0.4 MW.
1 1 1 1 1 1 1 1 1
___ 1_ _ _ _ _1 _ _ _ _ ..l _ _ _ _ L _ _ _ _ 1_ _ _ _ _ 1_ _ _ _ ....J _ _ _ _ J.. ____ L _ _ _ _ 1_ _ _ _ _ 1_ _ _ _
0.7 1 1 1 1 1 1 1 1 1 1 1
0.6
1 1 1
~ 0.5 _ _ _ 1_ _ _ _ _1 _ _
1
_ _ _ ___ L ____
1
1_ _ _ _ _ 1_ _ _ _
1
- - - - 1- _ _ _ _ _ _ _ _
1
::2: 1 1
O~
a..
0.4 I-----~~.---~----~--------~----.---~--~~--------~-.--~
'5 1
a. 1 1
____ L ____ L _______ _
1
____ L ____ L _______ _
'5 0.3 - - _ _ _ _ _ _1 _ _ _ -
1 1 1 1 1 1 1
o 1 1 1
1
1
1 1 1 1 1 1 1 1
0.2 - - -1- - - - -1- -
1 1
- - +- - - -
1
-1- - - - -1- - - - -1- - - - -1- - - -
1 1 1
+- - - - 1- - - - -1- -
1
- - -1- - - -
1
1 1 1 1 1 1 1
1 1 1
1
1
0.1 - - -1- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - T- - - - 1 - - - - - - - - - - - - -
1
5 10 15 20 25 30 35 40 45 50 55 60
lime [s]
Fig. 5.15: Turbine output power with storage (DFIG average model, 252 kW VRB)
Figure 5.16 shows the SOC variations during this simulation, while figure 5.17
shows the battery output power. The SOC fluctuates around its initial value of 50%.
When a wind storage system is designed, the expected wind variations in a

specifie location should be analyzed. The wind and battery systems should subsequently
be sized accordingly in order to ensure that the battery will not be under-charged or
overcharged over a long-term period.
-74 -
50.1
-+ __ -
50.08 - - -l - - _ - f- - - _ ...1- - - - _!_ - - - -1- ___ .j... _ _ _ --l ____ 1- ___ --l _ _ _ _ 1- __ _
1 1 1 1
__ ..J ____ L _ _ _ .1 _ _ _ _ 1_ _ _ _ .1 ____ 1_ _ _ _ 1.. _ _ _ ..J ____ L _ _ _ ..J _ _ _ _ L _ _ _
50.06 1 1 1 1 1 1
1 1
1 1 1 1 1 1 1 1
__ ..J ____ L _ _ _ .1 ____ 1_ _ _ _ 1.. ____ 1_ _ _ _ 1. ___ .J ____ L _ _ _ -' _ _ _ _ 1_ _ _ _
50.04 1 1 1 1 1 1 1 1 1
1 1 1 1 1 1
1 1 1 1
____ l ___ J ____ L _______ L __ _
50.02 1 1 1 1
~
Ü 50
0 1
CI)
1 l i t 1 1 1 1
49.98 - - -,- - - - , - --"T - - - -,- - - - T - - - - , - - - - T - - - 1 - - - - r - - - -, - - --r-
I l ' 1 1 1 1
1 1
49.96 - - ï - - - - r - - - ï - - - -1- - - - Î - - - -1- - - - T - - - Î - - - - r - - - Î - - - -1 - --

1 1 1 1 1
49.94 - - -J - - - - r - - - -t - - - -1- - - - +- - - -1- - - - T - - - -i - - - - t- - - -""i - - - - t- - - -

1 1 1 1 1
1
1
49.92 - - --1- - - - f- - - - +-- - -1- - - - +- - - -1- - - - +- - - --1- - - - f- - - - -1- - - -1- - - -
49.9
5 10 15 20 25 30 35 40 45 50 55 60
lime [s]
Fig. 5.16: Battery SOC (DFIG average mode), 252 kW VRB)

5
X 10
1.5,--,---,----,---,---,,---,--~----,_--,_--,_--_r--_,
__ ~ ____ L ___ ~ ____ L ___ ~ ____ ~ ____ L ___ ~ ____ L ___ J __ _

1 1 1
~ 0.5 __ ~ ____ L ___

1
~
1
____ L ___ ~ ____ ~
~
1
1
iL
~ 0
a..
è:'
2 ____
1
____ L ___ ____ L ___ J _ _ _ _ L __ _
êti -0.5 ~ ~ ~
III
__ -1 _ _ _ _ L _ _ _ ..J _ _ _ _ L _ _ _ _ 1_ _ _ _ .1 _ _ _ _ 1_ _ _ _ ..L _ _ _ _ 1_ _ _ _ ..1 _ _ _ _ L __ _

-1 1 1 1 1 1
5 10 15 20 25 30 35 40 45 50 55 60
lime [s]
Fig. 5.17: Storage system power response (DFIG average mode), 252 kW VRB)
- 75 -
The average DFIG wind storage model developed here is weIl suited for
medium term simulations, and can be used in order to validate complex control schemes
under various operating conditions.
In the next section, a real-time simulation of a similar system is considered,

which mixes simulated and experimental circuit components.
5.5 DFIG Simple Model for Real-Time Simulation
5.5.1 Centralized storage
A centralized storage scheme consists of directly connecting the storage system

to the grid bus, by using a dedicated ac-dc rectifier. In this case, the same storage system
can then be used for several wind turbines. This design is not limited to DFIG systems.
Again, the objective here is to make PGrid constant by controlling the power
flowing in and out of the storage device. This setup will be implemented in a real-time
simulation setup using a simplified DFIG model.
5.5.2 Real-Time System Implementation
Real-Time simulators can be used in various ways in order to help develop and
validate battery models, as well as obtain experimental results. Various RT platforms
exist, however the OPAL-RT simulator will be used, as it offers the best tradeoffbetween
complexity, accuracy and accessibility for this project.
Through Opal-RT's IIO modules, the system will include a mix ofreal elements
and simulated elements. The system will interface a simulated battery model, wind
system and controller with real converters.
5.5.3 Experimental Setup
Figure 5.18 below shows the real-time experimental setup implemented. The
Opal-RT IIO capabilities enable communication between the software and hardware. The
- 76-
real parts of the circuit include a one-phase voltage source representing the grid, as well
as an ac-dc converter, a dc-dc chopper and a controllable dc voltage source.
The role of the controllable dc voltage source is to act as the VRB in the real
world. The voltage is dependent on the State of Charge and cUITent, and is obtained from
the VRB model which is loaded in the Opal-RT simulator. The same approach could be
taken to represent various batteries in this way, simply by changing the model used.
Both converters are bidirectional. The dc-dc converter acts to set the cUITent,
based on the requirements to charge or discharge the battery. By loading the wind system
circuit in the Opal-RT simulator, the response of the storage system can be simulated
within a wind system. But first, the controis of the storage system need to be validated .
.DfgitalContrQller / AT
,:.,.,' '..
Simulator ;,. ""
Data acquisition and energy
, ' " ,
management control
Energy
Storage
Models
Icharge,
Idischarge
Pstorage
...
Vab
Controllable DC
Source
Hardware
Fig. 5.18: RTS experimental setup
The real circuit elements used will be operated in the 150 W range. However,
the simulated part of the circuit can be scaled up as needed, to represent the operation of
wind turbines in the MW range.
- 77-
5.5.4 Conffiguration and Validation of Control Systems
A simple PI control scheme is used, which sets the current based on a power
reference. Figures 5.19 and 5.20 show how the battery dc power follows the reference
c1osely.
5.5.5 Experimental Results
ln this case study, the same wind profile from section 5.4 is used. The wind data
is loaded into a IMW DFIG wind turbine simple model. Since this power fluctuates with
the wind, the purpose of the storage system is to smooth out the wind fluctuations. Figure
5.21 below shows the output of the wind turbine, without the storage device, and figure
5.22 shows the effect of the VRB system on the total output.
The power delivered to the grid is centered on 0.55MW. This is the scaled up
simulated power, while the actual power delivered in the circuit is in the 150 W range.
Figure 5.23 shows the power flow in and out of the battery model, and figure 5.24 shows
the VRB State of Charge fluctuations.
Figures 5.25 and 5.26 show the dc voltages and the dc current; the controlled dc
voltage source Vde follows c10sely the battery model reference Vbat. The dc bus voltage is
set at 300V. The dc current fluctuations correspond to the charge-discharge behaviour of
the storage device, as it tries to maintain the total output power of the wind system at a
constant value.
The developed RT system is well suited for battery model testing. In this case, it
was used to demonstrate the role of a VRB in a wind energy system. The design and
validation steps followed here can be repeated for various battery models as they become
available, in order to assess their performance.
- 78-
100 -------- - - - - - - -1- - - - - - - - - - r - - - - - - - - - -1- - - - - - - - -

1 1
1
~ 50 ------
....
Q)
3:
o
a.. o ----- ---~-----
Q)
o
c
~
j!2
cr: -50 -
Q)
-100 - - - - - - - - - 1" - - - - - - - - - -1- - - - - - - - - - r - - - - - - - - - -1 - - - - - - - - -

1 1 1 1
1
1
-150~------~------~--------~------~------~
10 20 30 40 50 60
Time [5]
Fig. 5.19: Storage system power reference (Centralized storage, RTS)
150~------~------~------~--------~----~
______ J __________ L _________ 1 ________ _

100 - - - - - -- 1 1 1
~ 50 - - - - --
....
Q)
~
a..
CI)
::J
III
g -50
_________ 1 _________ J ____________________ 1 ________ _

-100 1 1
-150~------~------~--------~------~------~
10 20 30 40 50 60
Time [5]
Fig. 5.20: Storage system response experimental results (Ceutralized storage, RTS)
- 79-
5
x10
10~------~------~------~------~------~
9 - - - - - - - - ~ - - - - - - - - - +- - - - - - - - - -1- - - - - - - - - -+ - - - - - - - -
1
1 1 1
8 - ~ - - - - t- - - - - - - - - -1- - - - - - - - - -t - - - - - - - -
1
1
.... 7
Cl>
~ 6
--
a..
::J
Q.
::J
5
o 4
Cl>
1::
~
1
::J
3 -------- --------- - - - - - - - -
1
-1- - - - - - - - - 1"1 - - - - - - - -
1- 1
1
J _______ _
2 ------------------ _ _ _ _ _ _ _ _ _ 1_ _ _ _ _ _ _ _ _
1 1
_ _ _ _ _ _ _ _ ..J _ _ _ _ _ _ _ _ _ L _ _ _ _ _ _ _ _ _ 1_ _ _ _ _ _ _ _ _ .1 _ _ _ _ _ _ _ _
~O 20 30 40 50 60
Time [8]
Fig. 5.21: Turbine output power without storage (Centralized storage, RTS)
5
x 10
10,-------,-------,--------,-------,-------,
1
1
1 1 1 1
9 --------~---------,---------T---------r--------
1 1 1 1
8 --------~---------,---------T---------r--------
[
1
1 1 1 1
7 --------~---------l---------T---------r--------
"'C 1
";:: 1
(!J 1
6
.9
....
Cl> 5 1
3: 1
--
1
a.. 4
________ ~
1
_________ ~
1
_________ L _________ L _______ _
1 1
::J
Q.
__________________ J _________ 1 _________ L _______ _
::J 3 1 1 1 1
0
1 1 1 1
________ _________ _________ L _________ L _______ _
2
~ ~
1 1 1
________ _________ _________ L _________ L _______ _

1
~ ~
1 1 1 1
1 1
1
~O 20 30 40 50 60
Time [8]
Fig. 5.22: Turbine output power with storage experimental results (Centralized storage,
RTS)
- 80-
2 -------- 1 1
---------r---------r--------
-
[1
::::s
.9-
::::s
o....
Q)
~
0-
~ -1 ------~---------,- -----r---------r-
> 1
1
1
1
1
1
1
1
1
1 1
1
-3L-------~------~~------~--------~------~
10 20 30 40 50 60
Time [s]
Fig. 5.23: Storage system power flow experimental results (Centralized storage, RTS)
1
1 1 1 1 1
_______ L _______ J ________ 1_ _ _ _ _ _ _ _ 1_ _ _ _ _ _ _ _ L _ _ _ _ _ _ _
50.08 1 1 1 1
1
1 1 1 1 1
_______ L _______ -1 ________ , ________ L _______ L ______ _
50.06 1 1 1 1 r
1 1 1 1 1
1
1 1 1 1
50.04 -------r-------l----------------r-------r-------
1
50.02 - - - - - - - -1 - - - - - - - -1 - - - - - - - -,- - - - - - - - - - - -
1 1
- - -
1
1 1
50 -------T-------i--
8 1
Cf) 49.98
49.96 ï
1
- - - - - - - -1- - - - - - - - r - - - - - - - r - - - - - - -
1 1
49.94 -------+-------4-------~--------~-------+-------
1 1 1 1 1
49.92 -------+-------4-------~--------~-------+-------
49.9
0 10 20 30 40 50 60
Time [s]
Fig. 5.24: Battery SOC (Centralized storage, RTS)
- 81 -
350~--~----~----~----~----~--~
300
250
~
~ 200
Cl
.s Ç;ontrolled :Voltage Spurce
g 150 - - - - - - - -1- - - - - - - - -1- - - - - - - - -1- - - - - - - - Î - - - - - - - - ""1 - - - - - - - -
ü
o 1
100 1
--------r-------------------------------------------
1 1
J ______ _
50 _ _ _ _ _ _ _ _ 1_ _ _ _ _ _ _ _ _ 1_ _ _ _ _ _ _ _ _ 1 _ _ _ _ _ _ _ _ _1 _ _ _ _ _ _ _ _
1 1
10 20 30 40 50 60
Time [s]
Fig. 5.25: De voltages experimental results (Centralized storage, RTS)
1~------~--------~--------~------~--------~
1
1
1 1 1 1
0.8 --------~---------,---------T---------r--------
0.6 ------~---------~--
1
1
1
0.4
-
~ 0.2
c
tg
:J
0
Ü
Ü
o -0.2
-0.4
1
1 1
-0.6 -----,---------r---------r--------
1
1
-0.8 ------i---------y---------r--------
1
1
-1
10 20 30 40 50 60
Time [s]
Fig. 5.26: De bus eorrent experimental results (Centralized storage, RTS)
- 82-
5.6 DFIG System Performance with VRB storage
5.6.1 Battery Energy Management
When a VRB is integrated in a power system, the controls must ensure that the
battery is not undercharged or overcharged, and remains between 20% and 80% SOC at
all times. However, additional factors have an impact on the battery energy management.
The power output reference of the wind system must be set appropriately so as
to ensure the SOC net power flow is zero over a long period of time, as the VRB storage
acts to smooth out the output. This was done in section 5.5 which resulted in acceptable
SOC fluctuations, as shown in figures 5.16 and 5.24.
Additionally, the battery must be sized properly in order to guarantee energy

availability for emergency situations. This requires the study of wind conditions on a
specific wind site, in order to obtain accurate forecasts. The stored energy can then be
managed accordingly.
The model implemented in the DFIG systems can be used to validate the sizing
of a wind turbine as well as the control strategies used, over a desired period of time. The
VRB system has proven to be well suited for this type of application, since the battery
energy capacity parameter can be easily sized, independently of other design parameters.
5.6.2 System Overall Efficiency
The system overall efficiency is dependent on the VRB efficiency (sections

3.5.2, 3.6.2), as well as the converter efficiencies (section 4.4.4). The VRB system
efficiency was determined to be a function of operating cUITent, as well as of the SOC.
Converter losses are typically of 4%. The battery dc round-trip efficiency,

including the dc-dc converter losses was found to be 68%. Additional converter losses
will bring the VRB storage system ac roundtrip efficiency down to 62.5%.
This efficiency compares well with other storage system, and makes the VRB
well suited for wind applications. However, the efficiency can drop considerably if the
- 83 -
control systems do not ensure optimized operation, which emphasizes the need for
complex control algorithms.
5.6.3 Impact on Grid
In sections 5.2, 5.3 and 5.4, the storage system ability to smooth the power
output of a wind turbine was demonstrated, using various DFIG system models, and on
various scales. This resulted in reduced power fluctuations, which justify the use of VRB
storage for DFIG systems, especially in weaker grids.
In remote hybrid diesel systems in particular, the storage system can be used to
allow shutting down diesel generators when they are running at low capacity, and thus
very low efficiency. If the wind suddenly drops, the VRB storage can then provide the
required power in order to allow the diesel system start-up, without power interruption to
the local grid, as was shown in section 5.3.3.
Battery storage is weIl suited for this type of application due to its quick
response as was shown in sections 5.3.2 and 5.4.2. The VRB in particular is unique in the
fact that it offers both a large energy capacity as weIl as a quick response.
5.7 Conclusions
In this chapter, the developed VRB model was used to demonstrate the
integration of VRB storage in DFIG systems, considering both local and centralized
storage schemes. This was done using three different DFIG models, each suited for a
different time frame. The storage system controls were presented and validated, and the
ability of the storage to smooth the output power of a turbine was demonstrated.
The VRB technology was shown to be an ideal choice for wind energy storage
applications, due to its large energy capacity and quick response, which allow it to be
used in both power smoothing (short-term) and load-Ievelling (long-term) applications.
The overall storage ac round-trip efficiency was found to be 61.5% including converter
losses, which is acceptable. However there is room for improvement. Efficiency was
- 84-
found to drop considerably under certain operating conditions, which emphasizes the
need for optimized control strategies.
The developed models can provide us insight on how to best operate the battery.
Based on the obtained results, recommendations for the design and operation of a VRB
will be presented in the next section.
- 85 -
Chapter 6: Conclusions and Future Work
Chapter 6
Conclusions and Future Work
6.1 Summary
6.1.1 Battery Models
The thesis focussed on the development of accurate and properly validated

generic models, for VRB and Li-Ion technology. The models took into account the
various battery physical and mathematical properties.
Various steps towards the validation of the models were discussed. A standard
charge-discharge profile for experimental battery testing, based on typical wind profiles,
was also presented.
The models allowed battery performance testing under varlOUS operating

conditions to determine:
• The VI transfer characteristic which provided the batteries operating ranges.
• The efficiency which was used to determine the best operating conditions.
• The transient response which enabled a better design of controller dynamics.
They were also used in:
• The design and validation of the storage size required.
• The design and validation of the PI control scheme used for power smoothing.
• The comparison of various integration schemes (centralized vs. local storage).
6.1.2 Integration into Wind Systems
The design steps for the sizing, control and integration of a VRB system in
various DFIG systems were presented. The key elements that needed to be considered for
- 86-
VRB sizing were identified, including the power rating and energy rating design. The
VRB model scalability was also considered. The scaled model provided the necessary
insight to test overall system designs.
The integration of the model in offline and real-time simulations of various

complexities allowed the testing and validation of the wind energy system design.
Various SIMULINK storage systems were developed including:
• Integration of battery on DFIG dc Bus, using DFIG detailed model obtained from
SimPowerSystems Library.
• Integration ofbattery on DFIG dc bus, using DFIG Average model obtained from
SimPowerSystems Library.
• Centralized storage scheme using battery model with a simplified DFIG model.
6.2 Conclusions
6.2.1 Modelling
The battery performance results matched the results obtained from alternative
sources, with a difference of less than 5%. The different system models are easily
modifiable and can be used as a basis for further studies related to VRB and Li-Ion
batteries, as weIl as other battery models, with the following conclusions:
• The detailed model is best suited for short-term simulations.
• The average model is best suited for medium-term simulations.
• The simple model is best suited for long-term simulations as weIl as real-Time
simulations.
The storage system power smoothing capacity in short-term (seconds) and

medium term applications (minutes) was also demonstrated using the models.
- 87-
6.2.2 Operating Considerations
Models such as the ones implemented and studied in this report provide us with
insight on how to operate batteries in an optimal way. Based on the results obtained in the
thesis, the following recommendations should be taken into account for operating battery
systems in wind energy applications:
• The battery should be operated at near-rated power whenever possible. The

control algorithm should take into account the efficiency data.
• For Vanadium-Redox batteries, the most efficient region of operation is between

0.7 and 1 p.u. charge or discharge power, at aState of Charge of 40% to 80%.
• A VRB should be operated between a SOC of 20% and 80%, where the voltage
is linear. This avoids overcharge and undercharge issues. The battery energy
capacity should thus be sized appropriately during the design.
• A Li-Ion battery should be operated at rated power or lower. Higher CUITents

result in reduced efficiency. It should also be operated at SOC between 20% and
80%.
6.2.3 Design Considerations
The thesis demonstrated that batteries are suitable for wind energy storage in
comparison to other storage technologies for the following reasons:
• High efficiency when operated properly compared to other technologies (Round

trip dc efficiencies before converter losses: VRB: 73%, Li-Ion: 92%).
• Fast response for power smoothing and load levelling applications (Worse case
response time: VRB: 52 ms, Li-Ion: 19 ms).
• Flexibility in terms of operating conditions: can be operated at a fraction of the

rated cUITent, rated cUITent, as weIl as maximum CUITent (up to 200% the rated
cUITent).
- 88 -
The Li-Ion battery showed better technical performance with a quicker response
and better efficiency. However the VRB technology appeared to be more effective than
Li-Ion for high-power applications for the following reasons:
• Quick response time (52 ms): suitable for power smoothing (short-term)
applications (higher frequency wind fluctuations are filtered out by wind
turbine).
• Energy capacity dependent on tank size and amount of electrolyte: well suited
for load levelling (medium and long-term) applications.
• Power rating dependent on number of cell stacks: well suited for high power
applications.
• High scalability: independence of power and energy ratings.
The following recommendations should be taken into account when designing

storage systems in wind energy applications:
• The choice between centralized and local storage depends on the system. A
localized scheme is justified for single turbines. A centralized scheme on the
other hand is advantageous for a wind farm.
• The battery sizing should account for the amount of energy that needs to be
displaced, the mean power at which the system should be operated, as weIl as
the power needed in emergency situations.
• A typical wind profile should be used as a standard to testing various battery

models, wind storage systems and control algorithms. Through having a
standard wind profile which represents the various possible wind conditions,
meaningful comparisons of various technologies can be obtained.
- 89-
6.2.4 A Methodology for the Study of Wind Storage Systems
Throughout this project, the various steps towards the study ofbattery storage in
wind energy applications were discussed. Figure 6.1 summarizes the various steps taken,
as a methodology for future studies. The emphasis is put on the development of battery
models and their use in these types of studies.
6.3 Future Research
This report has covered the modelling and validation of VRB and Li-Ion
batteries for wind energy applications. Sorne properties, like capacity fading, battery
cycling, heating and concentration depletion were not included in the models. These
battery characteristics should be taken into account in worse case operation analysis.
While they have no major impact on VRB performance, they should be integrated in
improved Li-Ion models, as weIl as other battery models. The battery energy capacity
design, which will impact the load leveling properties of the battery, should also be
considered with more detail.
Battery models, such as the ones developed in this thesis, allow researchers to
test systems in a fast and accurate matter, which reduces the need for on-site testing until
the final stages of a design proj ect. They provide the basis for developing smart battery
operation management algorithms. These algorithms should maximize battery efficiency
as weIl as battery life. The project economics, wind forecasts as weIl as application
requirements need to also be taken into account.
- 90-
DEVElOPING BATTERY MODEl
- Determining model needs

- looking at existing models
- Obtaining battery performance Data
BATTERY MODEl VALIDATION
- Obtaining Battery Performance Data

- Comparing mathematical and experimental
results using standardized tests
WIND STORAGE SYSTEM DESIGN
- Problem definition
- Battery sizing and control system design
- Short and long-term tests using
standardized testing procedure
- Overall system performance assessment
/ ~
DFIG DETAllED MODEl
- Battery model Integration
using local Storage Scheme
- Short-term simulations
l
DFIG AVERAGE MODEl
,---------------,
DFIG SIMPLE MODEl
- Battery modellntegration using

centralized Storage Scheme
(Seconds) - Battery model Integration using
- ledium and long- term
- Highly Accurate but slow and local Storage Scheme
simulations (Minutes - Hours)
computationally heavy - Short and medium term
- Fast simulation weil suited for
simulations (Seconds - Minutes)
long-term and real-time
- Tradeoff between accuracy
applications
and simulation speed
1 /
DEVElOPING SMART CONTROL
STRATEGIES
- Using battery performance data gathered

from models
- Taking into account operating efficiency
and dynamic response
- Optimizing system controls by considering
various strategies (Iterative process)
- Optimizing technical and economical
performance of system
Fig. 6.1: Methodology for the study and design ofwind storage systems
- 91 -
References
[1] J. P. Barton and D. G. Infield, "Energy Storage and its use with intermittent renewable
energy", IEEE Transactions on Energy Conversion, Volume: 19, Number: 2, June 2004,
Page(s): 441-448.
[2] M. Korpaas, A. Ho1en, and R. Hildrum, "Operation and sizing of energy storage for wind
power plants in a market system", Journal of Electrical Power Energy Systems, Volume:
25, Page(s): 599-606,2003.
[3] J. Paatero and P. Lund, "Effect of energy storage on variations in wind power," Journal of
Wind Energy, Volume 8, Page(s): 421--441, 2005.
[4] L. Leclercq, C. Saudemont, B. Robyns, G. Cimuca, and M. Radulescu, "Flywheel energy

storage system to improve the integration of wind generators into a grid", Journal of
Electromotion, Volume 10, Number 3--4,2003.
[5] C. Abbey and G. Joos, "A Doubly-Fed Induction Machine and Energy Storage System for
Wind Power Generation", Canadian Conference on Electrical and Computer Engineering,
Volume: 2, Page(s): 1059-1062, May 2004.
[6] C. Abbey and G. Joos, "Integration of Energy Storage with a Doubly-Fed Induction
Machine for Wind Power Applications" Power Electronics Specialists Conference, IEEE
35th Annual, Volume: 3, Page(s): 1964-1968, June 2004.
[7] E. Muljadi and J. T. Bialasiewicz, "Hybrid Power System with a Controlled Energy
Storage", Industrial Electrpnics Society, The 29 th Annual Conference of the IEEE, Volume:
2, Page(s): 1296-1301, November 2003.
[8] F. Zhou, G. Joos et al. "Use of large Capacity SMES to improve the Power Quality and
Stability of Wind Farms", IEEE Power Engineering Society General meeting, Volume: 2,
Page(s): 2025-2030, June 2004.
[9] Brad Roberts and Jim McDowall, "Commercial Successes in Power Storage: Advances in
Power Electronics and Battery Application Yield New Opportunities", IEEE Power and
Energy Magazine, Volume: 3, Number: 2, March 2005.
[10] A. Shibata and K. Sato, "Development of a Vanadium Redox Flow Battery for Electricity
Storage", IEEE Power Engineering Journal, Volume: 13, Issue: 3, Page(s): 130-135, June
1999.
[11] G. Koshimizu and Tetsuo Numata, "Applications of Energy Storage for Stabilization of
Wind Power in Power Systems", Subaru Project, ESA 2005 Annual Meeting Presentation,
May 24,2005.
[12] J.M. Hawkins and T.P. Robbins, "A field trial of a Vanadium Energy Storage System",
INTELEC 2001 Twenty-Third International Telecommunications Energy Conference,
Page(s): 652 - 656, October 2001.
- 92-
[13] Chad Abbey and G. Joos., "Short-Term Energy Storage for Wind Energy Applications",
Industry Applications Conference, Volume: 3, Page(s): 2035-2042, October 2005.
[14] A. Joseph and M. Shahidehpour, "Battery Storage Systems in Electric Power Systems",
IEEE Power Engineering Society General meeting, Page(s): 8, June 2006.
[15] H. Takabayashi et al., "Development of valve-regulated lead-acid batteries for Power

Storage", The Twenty-Fifth International Telecommunications Energy Conference,
Page(s): 383-390, October 2003.
[16] Z. M. Salameh, M. A. Casacca and W. A. Lynch, "A Mathematical Model for Lead-Acid
Batteries", IEEE Transactions on Energy Conversion, Volume: 7, Number: 1, Page(s): 93-
98, March 1992.
[17] M. Gatrell and N. Gupta, "Developing a simple flow battery model - Initial concepts",
Electricity Storage Systems Project, National Research Council Canada (NRC) document.
[18] S Mikaye and N. Tokuda, "Vanadium-Redox Flow Battery for a Variety of Applications",
IEEE Power Engineering Society Summer Meeting, Volume: 1, Page(s): 450-451, 2001.
[19] N. Miyanaga, T. Inoue, et al., "Large-Scale Li-Ion Battery Cells for space use", The
Twenty-Fifth International Telecommunications Energy Conference, Page(s): 241-248,
October 2003.
[20] I. Suzuki, T. Shizuki and K. Nishiyama, "High Power Li-Ion Battery for Backup Power
Sources", The Twenty-Fifth International Telecommunications Energy Conference,
Page(s): 317-322, October 2003.
[21] Ken Mattern, "Controlling Power Flow of Wind Farms", Engineering Power Quality
Products Division of S&C Electric Company, ESA 2005 Annual Meeting Presentation,
May 24, 2005.
[22] R. Rao, S. Vrudhula, D. N. Rakhmatov, "Battery Modeling for Energy-Aware System

Design", IEEE Computer Magazine, Volume: 3, Page(s): 77-87, December 2003.
[23] Marc Doyle, John Newman et al., "Comparison of Modeling Predictions with Experimental
Data from Plastic Lithium Ion Cells" Journal of Electrochemical Society, Volume: 143,
Issue 6, Page(s): 1890-1903, June 1996.
[24] S. Gold, "A PSPICE Macromodel for Lithium-Ion Batteries", Twelfth Annual Battery
Conference on Applications and Advances, Page(s): 215-222, January 1997.
[25] S. Buller, M. Thele et al. "Impedance-Based Simulation Models of Supercapacitors and Li-
Ion Batteries for Power Electronic Applications", IEEE Transactions on Industry
Applications, Volume: 41, Number: 3, Page(s): 742-747, May 2005.
[26] M. C. Glass, "Battery Electrochemical NonlinearlDynamic Spice Model", Proceedings of

the 31st Intersociety Energy Conversion Engineering Conference, Volume: 1, Page(s): 292-
297, August 1996.
- 93 -
[27] L. Gao, S. Liu and R. Dougal, "Dynamic Lithium-Ion Battery Model for System
Simulation", IEEE Transactions on Components and Packaging Technologies, Volume: 25,
Number: 3, Page(s): 495-505, September 2002.
[28] A. Shibata, S. Kanji et al., "Development of Vanadium Redox Flow Battery for
Photovoltaic Generation System", Conference Record of the Twenty-Fourth IEEE
Photovoltaic Specialists Conference, Volume: 1, Page(s): 950-953, December 1994.
[29] M. -H. Li, T. Funaki and T. Hikihara, "A Study of Output Terminal Voltage Modeling for
Redox Flow Battery Based on Charge and Discharge Experiments", The Fourth Power
Conversion Conference, April 2007.
[30] N. Miyanaga, T. Inoue, et al., "Large-Scale Li-Ion Battery Cells for space use", The
Twenty-Fifth International Telecommunications Energy Conference, Page(s): 241-248,
October 2003.
[31] I. Suzuki, T. Shizuki and K. Nishiyama, "High Power Li-Ion Battery for Backup Power
Sources", The Twenty-Fifth International Telecommunications Energy Conference,
Page(s): 317-322, October 2003.
- 94-

TFC Membrane

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

TFC Membrane

Transféré par

Droits d'auteur :

Formats disponibles

VANADIUM-REDOX FLOW AND LITHIUM-ION BATTERY

MODELLING AND PERFORMANCE IN WIND ENERGY

Thesis submitted to the McGill University Department of Electrical and Computer

Department of Electrical and Computer Engineering,

© John A. Chahwan, 2007

Published Heritage Direction du

Your file Votre référence

The author retains copyright L'auteur conserve la propriété du droit d'auteur

ln compliance with the Canadian Conformément à la loi canadienne

While these forms may be included Bien que ces formulaires

As wind energy penetration levels increase, there is a growing interest in using

La récente croissance en utilisation d'énergie éolienne augmente le besoin en

CHAPTER 1 INTRODUCTION •.•.•.•.•.•••••••••.•.•.••••••.•••••••••••••••..•.•••••••.•.••••••.••..•••••...•..•...•.•••• 1

1.1 Wind Energy Storage ........................................................................................... 1

CHAPTER 2 MODEL DEVELOPMENT ...............••.............•..............•...•.........•........•....•.•• 16

2.1 Introduction ........................................................................................................ 16

CHAPTER3 MODEL VALIDATION AND PERFORMANCE ............................................... 26

3.1 Introduction ........................................................................................................ 26

4.1 Introduction ........................................................................................................ 49

CHAPTER 5 WIND ENERGY STORAGE SYSTEM IMPLEMENTATION ............................ 63

5.1 Introduction ........................................................................................................ 63

CHAPTER 6 CONCLUSIONS AND FUTURE WORK •••••••••••••••••••••••••••••••••••••••••••••••••••••••••• 86

6.1 Summary ............................................................................................................ 86

1.1 Wind Energy Storage

The motivation for developing and implementing electrical equivalent models

• Short-term (seconds), medium-term (minutes) and possibly long-term (hours)

• Smoothing of large power output swings from wind farms;

• Meeting peak-Ioad demands without interruption and without Increase In

Wind is by nature unpredictable. However, In an electrical system, load

ln systems like Doubly-Fed Induction Generator (DFIG) wind turbines, energy

1.2 Storage Research

1.2.1 Types of Storage

Table 1: Available Power Storage Technologies

STORAGE ROUND TIME

Polluting (Uses fuel)

High flexibility Efficiency dependent on

Long cyclic life Short term storage

Storage options include Pumped Hydroelectric, Super-Capacitor Energy,

1.2.2 Common Batteries

Table 2: Main Battery Storage Technologies

Highly scalable, Cost effective,

High cost, Limited to lower

High cost and maintenance

Zinc-Bromine Minutes- Highly scalable, Medium and

1.2.3 Wind Storage Research

Vanadium-Redox Batteries (VRB) and Lithium-Ion (Li-Ion) batteries in

In order to study various aspects of battery storage in wind energy applications,

1.3 Battery Modelling

1.3.1 Battery Performance

Battery perfonnance needs to be evaluated when companng these vanous

• Discharge time, energy capacity and energy density

• Transient response time

• Rated and maximum power

• Cycle life and temperature effects

1.3.2 Battery Modelling Overview

Batteries have a series of characteristics that dictate their behaviour. These

Empirical models such as Peukert' s law [22] approximate a certain behavior of

Mixed models use high-Ievel representation of battery. The analytical

Abstract models provide an equivalent representation of the battery. The level of

1.3.3 Model Evaluation

• Type (physical, empirical, abstract, mixed)

• Variables of interest and total number of parameters

• Computational complexity and simulation time

PparaSitiC = P fued + k(~) = 84 + 42.s(~) (5)